AI at Work

You’ve collected 100 AI ideas. Now what?

Bryan Skwirut — Sun, 01 Nov 2026 13:26:37 GMT

Hey there 👋, I’m Bryan

Welcome to my newsletter. I share practical advice and tackle tough questions on putting AI to work. Thank’s for being here! Send me your questions and in return I’ll humbly offer actionable real-talk advice 🤜🤛

Subscribe now

In my earlier article on building an intake for AI projects, I covered how to get your arms around all the AI opportunities in your organization. You survey your employees and, best-case scenario, you get a great response. The next challenge is deciding which opportunities deserve your company’s time, money, and attention.

Your marketing team wants to automate a workflow. A vendor has a new feature that sounds promising. An executive wants to build an enterprise “brain.” Someone simply wants help drafting emails.

Now all of these projects are on the same list, waiting for you to decide what happens next. The trap? A hundred ideas can easily become a hundred distractions if you don’t know how to evaluate and prioritize them.

“Am I allowed to experiment?”

This is the first question your employees need answered. If the answer requires a meeting with three departments, you’ve probably made everyday experimentation too difficult.

The trick is to implement enough governance to make good investment decisions while leaving room for people to experiment and innovate. An employee organizing their meeting notes shouldn’t be subject to the same approval process as a team connecting AI tools to your customer database. Your process needs to make that distinction.

I would give employees permission to explore when all of the following are true:

They’re using an approved tool within its permitted uses.
They aren’t using sensitive or protected data.
The experiment stays within their own work and doesn’t affect a shared process.
It requires no additional spending or dedicated resources from another team.
The output is internal, and a human checks it before using it.
The tool cannot independently send messages, change shared records, or take other consequential actions.

Anything outside those boundaries should have a clear review process. That includes new tools, additional spending, engineering support, customer-facing uses, sensitive data, and changes that affect other people’s work. For example, an employee using an approved assistant to organize their own non-sensitive notes could fall comfortably within those boundaries. Connecting that assistant to a shared system and having it assign work to colleagues would require approval.

This first step should take a meaningful portion of your inventory out of the approval queue. Those ideas should still remain visible so others can learn from them, but they don’t need a committee’s attention.

Understand what kind of work you’re funding

Once you’ve separated everyday experimentation from initiatives that need review, organize the remaining ideas into a few useful categories.

I’d start with four:

Work with AI: Help employees do their work, such as preparing presentations, researching questions, or drafting documents.
Offer with AI: Improve the products and services customers receive, such as adding an AI assistant to an existing service.
Operate with AI: Improve how the organization runs, such as coordinating an intake process or automating a shared workflow.
Build with AI: Help technical teams design, develop, test, and maintain software and other technical solutions.

Some initiatives will fit more than one category. As a general guideline, put each initiative in the category that best reflects its primary purpose, and move on. The categories are there to help you make decisions, not slow down your process.

Then add a second dimension: experiment, pilot, or scale. These describe the stages of an initiative and highlight what you are being asked to approve next. Here’s a set of examples of how the two dimensions come together:

Worth noting: initiatives in the experiment stage here can still require formal approval. Using sensitive information or connecting to a live system doesn’t make it harmless just because someone calls it a test. Likewise, an expensive engineering pilot may carry relatively little privacy risk.

The category tells you where you’re investing, while the stage tells you what you’re committing to next. The risks determine which safeguards and reviews are necessary.

Decide what your organization values

Questions to consider before anyone starts scoring initiatives:

What are your organization’s strategic goals?
Does your team agree on the outcomes your entire AI portfolio should support and the balance of investment you want across the four categories?
How much capacity should go toward immediate productivity gains?
What resources are available to support new customer offerings or changes to how the business operates?

Your organization’s strategy may be focused on growth, customer experience, operating costs, clinical outcomes, something very specific, or something else entirely. This is generally a moving target than shifts year to year. Those priorities should shape how you evaluate AI initiatives, not the other way around. You may receive dozens of suggestions for personal productivity and very few for a difficult enterprise problem that deserves major resource investment and attention.

I like a weighted scorecard because it makes your priorities explicit. Here’s a starting point:

The first three criteria describe the opportunity’s alignment and value. The remaining four describe what resources it will take to deliver.

These weights give the criterion related to strategic alignment and business value a combined 45% of the score, and that’s intentional. Even so, the delivery criteria account for roughly half the total, which can favor inexpensive, fast, straightforward projects. Test whether the resulting rankings reflect your strategy. Otherwise, you may fill your portfolio with small improvements while repeatedly postponing the work that could transform the business.

Using the scorecard

Each criteria on the scorecard gets a value between 1 and 5, but before you start tossing out numbers, define with your team what each rating means. For example, a business value rating of 1 could mean the benefit is below a certain threshold or not yet measurable. A rating of 3 might mean a meaningful, measurable improvement to a single department or workflow. A score of 5 could then mean substantial financial benefit across the enterprise. Once you’ve agreed on rating definitions it’s time for the fun part.

Let’s get scoring! Rate each criterion from 1 to 5, with 5 representing the most favorable result. An easy way to remember this is that high business value earns a high score. High cost does not. An initiative that requires less investment should score more favorably on investment required, and an initiative with manageable risks should score more favorably on risk.

Optional enhancements to your process:

As an exercise to test your scorecard, take 3 to 5 real initiatives and score them against each other. Where people disagree, ask why. You’ll quickly discover whether the definitions are clear enough to use.
Request a short explanation and supporting evidence for each rating. Also, consider including a confidence measure as an additional dimension. A measured result from a pilot deserves different confidence than a vendor’s promise.

Time to calculate! Each criterion’s weight determines its available points. Business value, for example, has 25% weight and can contribute 25 points. Divide the rating by five, then multiply by the available points for that criterion. For example, a business-value rating of 4 earns 20 of the available 25 points. Add the seven results for a total out of 100.

The example below shows an initiative with strong strategic alignment and business value, but substantial investment requirements.

Hot tips 🔥

Calibrate investment thresholds to your organization and use a consistent cost horizon. A $100,000 initiative means something different to a small business than it does to a national enterprise.
Remember to consider vendor fees, internal labor, integration, infrastructure, maintenance, and ongoing usage costs.
Account time for training, workflow redesign, and the effort required to get people using the solution. An inexpensive tool can still require an expensive organizational change.
For time to value, define what a meaningful benefit looks like and estimate how long it takes to produce. Launching a pilot is a milestone, but it doesn’t establish that anyone’s work has improved, or that those improvements have showed up in the bottom line.

Once the rubric has been calibrated, keep it consistent for that prioritization cycle, and revisit it deliberately as strategy and experience change.

As a final note, keep mandatory approval requirements outside the score. Clinical safety, privacy, security, regulatory compliance, and vendor due diligence deserve their own reviews. This prevents an initiative from offsetting unacceptable risks by earning extra points elsewhere.

Decide when you’ll stop before you start

The scoring is complete, and now you have a ranking system to help improve the quality of your decisions. Hooray! This next part is often overlooked.

Most of us have experienced working on a project far longer than we should have, maybe even knowing it was a dead end. This is because organizations spend a lot of time defining how projects get approved and less time defining when projects should end. Planning for the end is not a silver bullet, but it can help ease the tension that often accompanies shutting down a project (letting go is hard!).

Before an experiment or pilot begins, agree on:

What it must demonstrate.
How you’ll measure success.
Resources you’re willing to invest.
Which problems would require an immediate pause.
What is your review date, and who will decide whether to expand, revise, or stop the work.

Suppose an AI tool reduces the time employees spend drafting something, but increases the time others spend checking and correcting it. You need to know whether the overall workflow improved. Measuring only the drafting step could give you a very flattering answer to the wrong question.

When the review date arrives, ask whether the evidence justifies the next investment. If it doesn’t, stopping is a useful outcome. You’ve learned something while limiting the commitment, and you can put those resources toward a more promising idea.

That decision gets harder after people have invested months of effort and attached their reputations to the project. Agreeing on the criteria upfront gives everyone a fairer basis for the conversation.

Recap: What will you commit to?

Remember:

A useful AI portfolio should help you explain why you’re funding one initiative, why another needs more evidence, and why a third can proceed without anyone waiting for approval.
Employees should know where they have permission to explore. Teams should know what it takes to earn investment. Leaders should know what evidence would make them change course.

Now when your next meeting opens with a list of 100 AI ideas, you’re prepared to ask: what do we value as an organization? which of these ideas are we prepared to commit people and resources to, and what do we need to learn before committing more?

That’s the conversation that turns a growing inventory into strategic decisions.

Thanks for reading!

Sincerely,
Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

Your AI policy isn’t working if no one follows it.

Bryan Skwirut — Sun, 18 Oct 2026 12:31:35 GMT

Hey there 👋, I’m Bryan

Subscribe now

Does your organization already have an AI use policy? If so, congratulations! Less than half of organizations and businesses have one. But having a policy doesn’t necessarily mean it’s doing its job. Whether you’re writing one from scratch or wondering if your existing policy is effective, here is what I’ve learned about writing an AI policy your team will actually follow (jump to the six steps).

Years before anyone could paste sensitive customer data into a chatbot, I spent what felt like a lifetime (ok, just a year) under the fluorescent lights of a 1950s office-building basement working on a market conduct exam for a big insurance company (Backrooms, anyone?). If you’ve never had the pleasure, a market conduct exam is when insurance regulators pull every business decision you’ve ever made and check it against the rules and the procedures you created and swore you were following. Ours went back 10 years.

Insurance is a unique industry where the policy is literally the product, and even so, most policies go unread by the very customers who are paying for them. This is where examiners come in.

Turns out, having the right procedures on paper doesn’t count for much. The examiners demanded proof we had actually followed them, file after file, and the stakes were quite high. Another big insurer came out the other end of an exam like ours with a $19.75 million fine and more than $189 million in payback payments. We had this in the back of our minds as we worked diligently in that basement. It was the cautionary tale everyone on our team was quietly trying not to become.

I think about that basement a lot now in my AI work, because the examiner’s question matters for an AI policy too: do people’s actions actually match what the policy says?

Most policies go unread. Why?

Be honest, when was the last time you read the terms and conditions for your latest app update?

The legal system rewards companies for having policies, making it tempting to write one for the auditor and call it a day. Legal and compliance usually hold the pen, and they’re tasked with avoiding liability, so they draft the document to hold up in a deposition rather than to be read by the average Joe on a Tuesday morning. Most policy sign-offs only check that you clicked “I acknowledge,” so that’s what gets optimized.

And every time a new issue arises, compliance adds a clause. But nobody ever takes one out. A lot of policies are like scar tissue from an isolated incident, and the policy outlasts everyone who remembers why it’s there.

That’s how you end up with policies people quietly work around. Most policy violations are workarounds for a real problem, and if you ban the workaround while leaving the problem in place, people will just route around it.

The good news is that the auditor is getting harder to fool. Regulators increasingly want proof that a policy actually changed behavior, which is exactly what those examiners in the basement were digging for, file after file. So the best legal defense and the best adoption strategy are turning into the same thing: a policy people actually follow.

Here’s what makes an AI use policy difficult to write

An AI use policy might sit in the same handbook as the travel and expense policy, but it’s far more complex.

First, your people are already using AI, with or without your blessing. A huge survey of UK workers this year found nearly two-thirds use it for work, and about a third of those users haven’t mentioned it to their boss. By the time your policy shows up, it’s walking into a party that started hours ago, mostly on personal accounts. And being fashionably late here is definitely not cool.
Second, your go-to fix of reminding everyone about the risks doesn’t do much. In one recent survey, plenty of healthcare and finance workers were using unapproved AI tools even though most of them said they understood the risks. Nearly half of IT workers said they’d keep using those tools even if their employer banned them and threatened discipline. These people read the warning label and decided the tool was worth it anyway.
Third, you’re writing rules about a quickly changing landscape. The share of big companies reporting serious employee adoption of AI roughly doubled in a single quarter, and a lot of them are now building AI agents. A policy written for a chatbot that drafts emails doesn’t apply to an agent that’s now sending them.

A simple takeaway is that a “ban-first” policy doesn’t stand much of a chance. It sends your employees to their personal phones and accounts, where whoever owns the policy can’t see a thing. How do I know? This hidden use is exactly what’s been showing up in more and more breach reports. The job isn’t just to write rules that address today’s risks. It’s to write something people will actually choose to follow, even as the tools change.

Six steps to writing a useful AI policy

Remember, none of this is complicated. Follow these six steps to avoid accidentally writing an ineffective policy.

Find out how people use AI now. Do this before you write the policy.
1. Send a short, anonymous survey.
2. Talk to people in different roles.
3. Get data from IT about which AI sites people use.
4. Ask what worries them about AI.

Ask people which tools they use and what they use them for. Tell them that their answers will not get them in trouble. Each workaround you find helps you write your policy.

Write the policy with the people who must follow it. Make a small working group comprising a healthy mix of:
1. Your heaviest AI users.
2. One or two skeptics.
3. A frontline manager.
4. Legal, privacy, security, HR, and IT.

The group’s first task is to complete one sentence: “This policy keeps _____ safe from _____.” If the only answer is “the company from lawsuits,” start again. Legal will help set the minimum requirements, but the others will make it usable.

Give people an approved tool before you restrict anything. Before launch, have these items ready:
1. One or more enterprise tools with the necessary data protection.
2. A clear list of approved uses, in plain verbs (for example: summarize, draft, rewrite, translate).
3. A process to request new tools, with an established response time.

If the approved tool is worse than the workaround, people will still use the workaround. Last week I explained in Microsoft Copilot is the Frontdoor to AI why free tools and their lack of context are inferior in the workplace.

Set clear limits and explain the reason for each. Most AI policies need limits for these topics:
1. Which data can go into which tools (customer, employee, health and financial data).
2. Who is responsible for AI output (the user, which human being must check it).
3. When to disclose AI use.
4. Which decisions need human approval (for example, hiring, pay and patient care).
5. What AI agents can do without human approval.
6. What to do when you are not sure, and what to do after a mistake. Importantly, do not punish people for reporting their own mistakes. If you do, people stop reporting.

Give a reason for each limit. People who understand the reasoning are better equipped to make good decisions in new situations. Keep the policy short, and provide real examples when possible.

Separate stable rules from rules that will change. Write the policy in two sections:
1. The core: include main principles, data limits and accountability. This part should be evergreen and rarely need changing.
2. A living page: list of approved tools and use-case guidance. Plan to update this section on a published schedule, for example monthly or quarterly.

Major policy reversals cause confusion. A published update schedule can help prevent most reversals. Write the next review date and the owner’s name directly in the policy.

Test the policy before launch. Use 10 to 15 real scenarios from step 1. For example:
1. A recruiter summarizes resumes.
2. An analyst uses customer data in a spreadsheet.
3. A manager wants an agent to send follow-up emails.
4. A clinician writes a visit note.

If two people get widely different answers, change the policy. Then do a pilot with one or two teams for a few weeks. Add the hard scenarios and their answers to the policy as examples.

Ready to launch it?

But wait! After all that work, it’s tempting to just email the PDF to the whole company and head out for lunch. Not so fast. How you introduce the policy matters just as much as how you write it.

A lot of damage can happen in the first few weeks after a policy goes out, and in my experience, it usually follows the same script: It was drafted without the right people and shipped before anyone tested it against real work. Six weeks later, a revised version is reissued because somebody added something obvious the right people would have caught earlier. By then, employees have possibly read one of two versions and heard three interpretations from their managers. What then? They’ve quietly gone back to whatever they were doing before. And then your policy just collects dust on the proverbial shelf.

My solution? Plan the release in tiers. Give your managers the policy at least a week early, with talking points and room to push back. Nothing sinks a policy faster than an employee asking their boss about it and finding out the boss is reading it for the first time, too.

When it’s finally ready to go out to everyone, lead with callouts of what people can use starting today, then the handful of lines they can’t cross and why each one is there. By the end of that week, everyone should know who to ask when they’re unsure, how to report a mistake without dreading the conversation, and when the next update is coming. If your managers can’t explain all of that in their own words, there is more work to be done.

In the following weeks and months, find ways to measure what people actually do, instead of counting who clicked “I agree.” Look at how many people are using the approved tools, whether traffic to unapproved AI sites is dropping, and how many mistakes get reported.

Early on, more reports are a good sign, because it means people trust the process enough to use it. Those numbers tell you whether the policy is working, and they’re exactly the evidence a regulator could ask for someday.

The one question to ask of any policy

When someone hands you a policy, AI or otherwise, one question tells you most of what you need to know about it: if I can’t do this, what am I supposed to do instead?

A policy that can answer that question was written by someone who understood the work. It points you to the approved tool, the person to call, or the process to follow when the usual way isn’t allowed. A policy that can’t answer it was written for the auditor. It tells you to stop solving a problem without telling you how else to solve it. Earlier I said most policy violations are workarounds for a real problem, and this question is the quickest way to find out whether a policy left one of those problems in place.

If you’re the one writing the policy, run every line of your draft through the same question. Anywhere the answer is “nothing” or “it depends,” you’ve found a gap worth fixing before the policy goes out.

The tools in your policy are going to change a lot over the next few years. As long as each version still answers that question, people will have a reason to keep following it. And if those examiners ever come knocking, you’ll have something better than a beautifully written document. You’ll have evidence that people actually use it.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

Microsoft Copilot is the digital front door to AI. What does that mean for your transformation strategy?

Bryan Skwirut — Sat, 10 Oct 2026 13:04:33 GMT

Hey there 👋, I’m Bryan

Subscribe now

This week I asked Copilot to build me an action tracker in Excel, and as someone who spends significant time with premium AI tools (too many $200/mo subscriptions!), I was not impressed.

Copilot created the action tracker, and then saved the file to OneDrive. A few prompts later, when I tried to edit the file, Copilot told me it no longer had access to it 🤔. It turns out, Copilot couldn’t access or modify the file it had created just a moment ago, and in an effort to be helpful, offered to create an entirely new file (version control, anyone?). All I had to do was download the original and upload it to the chat.

What the free Copilot couldn’t do

I can’t think of a clearer example that highlights how important context is for AI to work properly and provide real productivity gains, as opposed to wasting time and building frustration. Context is the data and tools an AI has access to in order to effectively complete a job.

This was the first of several frustrations with the free version. Not only could Copilot not see anything in my OneDrive, but Excel, PowerPoint, Word, and the rest of the O365 Suite have no Copilot integration on the free plan. Copilot is effectively on an island. It doesn’t know anything about the scope of my work, my company’s goals, or literally anything beyond its generic training data. Its limitations make it practically useless.

Compared to the premium AI tools I use every hour of every day, this was an extremely weak experience. But why?

I confirmed that I had a full Office license; however, the paid Microsoft 365 Copilot isn’t included in the package, and most companies don’t hand it out by default.

Copilot is the front door

Most mid-size and large companies run on either Microsoft or Google, and for the ones on Microsoft, Copilot is the digital front door to AI for employees. Microsoft 365 has more than 450 million paid business seats, but fewer than 7% provide paid Copilot to employees, which means a free AI tool is the cornerstone of many companies’ AI transformation strategy.

Here’s what I think leaders are missing. Even though the paid version runs on largely the same AI, it adds the secret sauce: context. Think of the free version as a personal assistant… on their first day. They might be highly skilled, but they won’t reach their full potential until they have access to your email, your files, your meeting notes, your team’s SharePoint sites, and the apps you already use.

My before and after

Back to my action tracker workbook in Excel. Here’s how that task went once the full license for Copilot Premium came through.

After a request to IT, Copilot showed up in Excel pretty much immediately, though I quickly noticed that I didn’t need to open Excel at all. In Copilot Cowork (a premium feature), I typed “/” to pull up the action tracker file, described the updates I wanted in plain language, and Copilot updated the file directly. One file, no copies. The same “/” pulled up every file I needed.

Then I took it a step further and asked Copilot to keep my action tracker updated from my meeting notes, Teams chats, and emails. That’s the whole difference in one example. The free version couldn’t open a file it had just created. The paid version keeps that file current from everywhere my work happens without me even needing to open it.

It helped in ways I didn’t expect, too. I could find company policies just by asking in chat. For most issues, I’ll never have to email someone and wait for an answer, saving company time for both me and the person I would have asked.

The biggest win was my calendar. I used Copilot to schedule all my meetings for the week. I loathe trying to find available times on people’s calendars, especially when I have multiple meetings to schedule. So I had Copilot do it for me. It found available time on calendars, sent invites, flagged when someone declined, and found an alternate time. That saved me well over an hour of calendar searching and back-and-forth.

You might be underselling AI to your people

If Copilot Chat is how your company is letting employees try AI, they are using the weakest version available.

Your employees aren’t fooled. Most of them can tell it’s underpowered. The trouble is what they conclude from the experience. Let’s look closely at three employees.

Employee A has never used AI outside of work. Since they have never tried a powerful version with full access to the necessary context, they may run into a frustration similar to the one I encountered with OneDrive files. They decide AI is overhyped and not very useful. They keep doing great work, but without the added benefits of a powerful tool. When leadership asks how the AI rollout is going, the answer is lukewarm.

Employee B uses paid AI tools outside of work and fully understands AI’s potential. When the approved company tool can’t keep up, the obvious move is to paste work into the tools they already like and use. In healthcare, finance, or anywhere with sensitive data, that’s a risk. Their chosen tool helps with their work, but it still isn’t integrated into the company ecosystem, meaning that if/when they leave your company, whatever agents or workflows they’ve created go with them.

Employee C understands the limitations (and risks) to using their own personal AI tools at work, and they demand better. Because you are limiting access to more powerful models, they are actively considering leaving your company to find an employer who will so they can improve their AI skills and stay relevant in a rapidly changing space.

The bigger your company, the weaker the free version

Since April 15, 2026, Microsoft has removed Copilot from Word, Excel, PowerPoint, and OneNote for employees without a paid license at organizations with more than 2,000 Microsoft 365 seats. Those employees still get Copilot Chat in the Copilot app and Copilot in Outlook, but the Copilot panel inside the apps where they actually work is gone. I would argue employees at the largest companies have the most to gain from AI. As leaders, keep in mind their feedback will be based on the thinnest, most watered-down version Microsoft offers.

The missed opportunity

If you’re a Microsoft shop, the beauty is that your people’s work already lives in a contained ecosystem: OneDrive, SharePoint, Excel, PowerPoint, Word, Outlook, and Teams. Paid Copilot is wired into all of it. As an experienced transformation leader, I cannot stress enough how valuable this is. You have no integration project ahead of you, no data migration, and you likely already have a business associate agreement with Microsoft that includes support for the enterprise controls you need, like data-loss prevention, prompt/response records, and retention.

Here’s what a connected Copilot might look like on any given day:

Excel: instantly build a formula for your workbook, analyze the data, and flag outliers.
Outlook: summarize a long thread, identify to-dos, and draft a reply that reflects what was already agreed upon.
Teams: recap a meeting with decisions and action items.
PowerPoint: turn a Word doc or a set of files into a first-draft deck.
SharePoint and OneDrive: ask questions across your team’s documents, and build pages and dashboards from lists and files (Copilot in SharePoint and OneDrive reached general availability starting September 30, 2026).

Big buyers are deploying widely: The University of Pittsburgh Medical Center is among the companies that have bought 60,000 or more seats. The National Health Service (NHS) England is extending Copilot to 505,000 clinicians and staff. Microsoft says Copilot’s weekly engagement is now on par with Outlook and Teams, and that customers now reach high usage within days of buying licenses. Being skilled with Copilot will become table stakes, the same way Excel proficiency is expected.

The cost is close to moot compared to productivity gains, but it can still be an initial hit to the bottom line

Microsoft 365 Copilot Premium lists at $30 per user per month, billed on an annual commitment, so $360 per seat annually. It sits on top of your base license. The Enterprise E3 Tier is the “corporate baseline” and costs $39, bringing the all-in cost to $69 per month. You pay for the base license either way, so the real decision is whether to spend the added $30.

What does your people’s time cost? The latest data from the Bureau of Labor Statistics puts total compensation for full-time private-sector workers at $54 per hour worked in June 2026. That’s wages plus benefits, or about $0.90 per minute.

$30 a month ÷ $0.90 a minute ≈ 33 minutes a month, or about 8 minutes a week.

A seat pays for itself if it saves an employee only 8 minutes of work per week, or under two minutes each workday. Averaging across all private-sector workers ($46.89 an hour), you could safely estimate around 9 minutes per week.

For comparison, NHS England reports its Copilot trial saved employees an average of 43 minutes a day. Even if your organization’s results land at a fraction of that, the math still works in your favor by a wide margin.

That doesn’t make the up-front costs cheap at scale. For a 10,000-person company, Premium for everyone is $3.6 million a year, paid on an annual commitment. Agents can add more: licensed staff using internal agents are mostly covered, but agents used by unlicensed employees or customers bill by usage, and that cost is hard to predict. This is a real commitment, and the savings come back as time.

But if you look at it like any other core tool, a company doesn’t decide whether to give employees Excel or a laptop based on efficiency alone. These days, it’s just what the work requires. AI is heading the same way, and fast. Treat it as an optional perk and your workforce will fall behind the companies that treat it as standard equipment.

How to ensure it pays off

If AI is becoming standard equipment, the question is how to make every seat earn its place. A major failure point in any internal AI rollout is when employees don’t use the tool. A good rule to follow to limit unnecessary costs: training unlocks spend, and proven value earns more. $30/mo is the base price, and you can unlock additional usage limits by paying more. That structure works well if you manage it purposefully:

Train before you assign. Require training on safe, efficient use as a condition for getting a license. Employee A needs to know what the tool can do before the seat can pay off.
Ask what it’s for. Use the license request as an opportunity to gather helpful information about your team, including what AI tools they use today, current use cases, and expected time saved.
Treat the usage limit as a conversation. When someone hits their usage cap, ask them what they’re working on and why the limit falls short. Most increases will be small and easy to approve, and that conversation is where you weigh cost against benefit.
Check back every 60 to 90 days. Ask each user to name one weekly task Copilot does for them. If they aren’t using Copilot regularly, follow up again or reassign the seat at renewal.
Set a company-wide cap. Individual increases should add up to a budget you chose, not one you discover.

Spend should grow only as capability and value grow with it.

So here’s the question for every CEO running on Microsoft: are you willing to commit real money to a tool whose cost is easy to see and whose return is spread across thousands of small moments? Companies that make that investment, then measure it, will learn what AI can actually do for them. The ones that don’t will keep judging AI by its weakest version.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

How to measure AI ROI beyond the token

Bryan Skwirut — Sun, 27 Sep 2026 20:46:57 GMT

Hey there 👋, I’m Bryan

Subscribe now

Not long ago, organizations incentivized software engineers by the volume of code they wrote. Unsurprisingly, when performance was judged on volume, the result was bloated repos, low-quality output, and engineers who looked busy while the codebase quietly decayed. In other words, organizations got exactly what they measured.

Organizations rolling out AI today are at risk of making a similar mistake. The AI adoption mandate is the modern equivalent of grading on lines of code. This time, however, AI lets anyone produce an abundance of material fast, which means anyone can produce lots of junk fast, then wave it around as proof of “adoption.” Once quality is out of the equation, all that’s left is activity. At worst, people start gaming the metrics.

When adoption becomes the goal
A growing part of everyone’s job is cleaning up the AI slop from people trying to operate outside their expertise. Shopify’s CEO, Tobias Lütke, calls these “slop grenades.” Yes, people have cleaned up work slop forever, and AI slop is just the newest flavor, but the difference now is that AI slop is hollow while presenting as confident. Oh, and there’s exponentially more of it, which makes it impossible to review everything.

In the race to look AI-forward, behaviors like “token-maxxing” and “budgeting for tokens” made their way to the C-suite and into the mainstream conversation. Amazon employees built an internal token-usage leaderboard that backfired, as people routed unnecessary work through AI to inflate their scores. Amazon shut it down, citing the compute costs it generated, and a senior VP told staff to stop using AI just for the sake of using AI. Many companies seeing similar behavior are responding with usage limits and token budgets.

The fallacy of budgeting for tokens
On the surface, budgeting for tokens makes sense. Tokens cost money, so it’s only logical to budget for them. If an organization’s end goal with AI is to hand its workforce a chatbot and throttle usage, then yes, budgeting for tokens is probably the right approach. Organizations with ambitions beyond adoption and chatbot usage will need to evolve their thinking to match how AI agents and automations actually show up in the enterprise.

Traditional software is budgeted by the seat, because its cost grows with the number of people using it. Chatbots are budgeted by the token, because their cost grows with how much people use them. Agents take action, whether that’s issuing a refund, updating the CRM, or processing a claim, so their cost grows with the work they do. A refund issued correctly is worth something, whereas a refund issued wrong costs you money, however few tokens used. That’s why the unit of measure has to move from the token to the task.

So when it comes to ROI, where does cost per token fit?
Cost per token is an input, and inputs make poor scorecards. Yes, every task is still made of tokens, and token price belongs in the math, but it shouldn’t be the number you manage to.

Imagine an insurance workflow
Instead of asking, “how much does this model cost per million tokens?” or “how many millions of tokens, and at what rate, did we pay last month?” ask, “how much does it cost to process one claim correctly?”

Then decompose the workflow into tasks:

Intake: cheap model
Document interpretation: stronger model
Fraud anomaly detection: specialized model
Decision recommendation: frontier model
High-stakes exception: human

Each step gets the level of intelligence it needs and no more. Now suppose someone decides to save money by moving document interpretation to a cheaper model. The token bill goes down, but the cheaper model might misread more documents, so more claims get flagged for human review. Because one human review can cost more than every model in the workflow combined, the token budget shows a win while the cost per claim goes up.

Where this is headed
Here’s what I’ve come to believe will define mature AI-native operations. Companies will stop standardizing on a single model, or even a single model provider, and workflow steps will be routed to the least expensive level of intelligence that still clears the quality bar. Microsoft is already building for this. Its Foundry model router reads each prompt and sends it to one of dozens of models from OpenAI, Anthropic, Meta, and others. Microsoft 365 Copilot now runs on models from both OpenAI and Anthropic, and this month Microsoft added the newest from each. Once routing is the norm, the price per token stops telling you much.

Humans already work this way
Nobody asks a VP to handle every administrative chore just because they’re more capable than a director. Similarly, AI-native organizations won’t run every task through their most expensive model either. Model routing is organizational design, and because agents take action, model choices become part of how the business operates.

Nobody measures an employee in words per dollar. So why would we measure AI that way?

Cost per token
↓
Cost per task
↓
Cost per workflow
↓
Cost per outcome

Each step down the stack gets closer to an actual business case for AI. A model that’s 3x more expensive per token can be cheaper per completed workflow if it uses fewer steps, needs fewer retries and requires less human cleanup.

How to start

Pick one workflow that ends in a clear outcome.
Find out what it costs today. Before any AI, what does one completed outcome cost in people’s time and tools? This is the baseline.
Break it into tasks and set a quality bar for each step.
Assign each task the right model (or person) that clears the quality bar.
Measure cost per completed outcome, including any human intervention, and compare it to the baseline. The difference is the real cost (or savings) of AI.

Beyond the token
Most organizations starting an AI transformation want to grow beyond chatbots and usage dashboards. This is a marathon, cost per token is the starting line, and cost per outcome is the finish. What organizations are really paying for is the outcome. Organizations that measure what they actually want may be surprised how much of the slop was never worth the tokens in the first place.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

Four frontier AI agents vs. one simple task

Bryan Skwirut — Thu, 17 Sep 2026 14:51:39 GMT

Hey there 👋, I’m Bryan

There are so many AI tools available right now, and it’s maddening to try to keep up with all the new product launches. SpaceXAI launches Grokbot and says they built it in a month. Meta’s Muse pops up in your Facebook messages. Hermes and ChatGPT Work are out there, and Open Claw appears to have quietly faded into the background.

So I decided to run one simple task across four tools. The task: buy a housewarming gift for a friend. I knew exactly what I wanted, a 4-quart round Dutch oven in basil with a brass knob, shipped directly to my friend, gift-wrapped with my own personal message.

I tried Grokbot, Claude, ChatGPT, and Muse with mixed results. And yes, I understand the irony. I’m spent hours trying to get an agent to buy a gift I could have bought myself in about two minutes. I wasn’t doing this just to buy a gift, though. I was doing it to learn and share with you.

Grokbot

I had just finished watching a full 2-day livestream of SpaceXAI employees using Grokbot to build their own business. I figured if they could do that, and I could spend 16+ hours watching them, then I could probably order a crockpot with the Grokbot.

Grotbot is so hot right now. It’s also brand new.

Despite yet another $200 monthly subscription to yet another AI tool, getting started was pretty easy. I set up a chief of staff as the main bot I talk to, and that bot could work with other bots on my behalf. That way, I’m not tracking every bot I have floating around.

Here’s the cool thing about Grokbot if you’ve never used it. To create a bot, you click a plus icon, tell it what to do, and it prompts you back and forth. You give it a name, an identify, and the tasks you want it to execute.

One pro tip I picked up from watching the XAI employees: there’s a bot marketplace. Inside it, you can use bots other people have already built. When you use someone else’s bot, you don’t get its memories, data, or conversation history, but you do get all of its capabilities. One bot, created by Lauren Tan, is called dr. eggbot. dr. eggbot’s superpower is creating other bots, very well. If you want to build new bots, you use Dr. Eggbot to build them.

So I used dr. eggbot to create Gifted Gary. Gary’s job is to buy gifts for my friends. Perfect, because I knew exactly the gift to buy.

Another cool thing about Grokbot is that you can watch the agent’s screen as it works. I could see Gary navigating around, trying to find the product. One of the selling points is that the bots don’t take control of your computer. They’re cloud agents with their own computer, and they use it to execute tasks on your behalf. That appealed to me, because I’ve never been comfortable giving agents full control of my machine. Pretty quickly, though, I realized that was a benefit in some areas and a blocker here, because William Sonoma’s website appears to block bots from crawling or using it.

Once Gary got blocked with its own computer. things started going downhill pretty quickly. My chief of staff stopped responding entirely. I quit and reloaded. I turned it off and turned it back on again. Still nothing. I checked my usage, and I still had 21% of my trial left. Come on, Grokbot.

At first I thought there might be an outage, but I was still able to talk to dr. eggbot, so I concluded it was only my main bot that had stopped working. I created a replacement chief of staff, Chief Charlie, and deleted the old one. You’re gone. Boom, poof. You’ve been replaced, and we were back up and running.

Getting past the blocker

Chief Charlie started working with Gary again, and Gary hit he same website blocker. Eventually, Grokbot asked to use my own Mac. That felt like a downside, having its own computer was the appeal, but if the website blocks that computer, it is what it is. After some trouble getting the permission approved, Grokbot opened Safari (wtf?), then multiple Chrome windows. It kept opening another browser instance and going to the merchant website, and some commands appeared to time out. It was doing things, but in a much more invasive way than I was comfortable with.

Eventually, Grokbot selected the right product and added it to the cart. It added it twice, then removed one. It added the gift wrap, and it looked like it might get there, but it did completely take over my computer. I couldn’t do any other work while it ran, which was a real downside.

Where I left it: Grokbot was working through the cart and gift options. I didn’t record a completed purchase. I still really like the idea of a team of specialized bots, but this attempt involved a frozen bot, a replacement bot, a blocked computer, and a fairly disruptive takeover of my Mac. I’ll try again in a few weeks.

Claude

I started by talking to Claude about the gift idea, and that part was helpful. I could talk through the product, learn more about it, and land on the specific one I wanted and why. Then I said, essentially, “Go ahead and purchase it. Do you need the address?”

Claude said it couldn’t make the purchase or check out anywhere, so no address was needed. It could, however, point me to where to buy it, draft a gift note, or set a reminder to order it.

My honest take is that this is horribly unhelpful. The last thing I wanted was for it to write the gift note. I wanted that part to be personal, and I didn’t need a reminder handing me another thing to do later. That’s what frustrated me. It could take away the human part, or it could give me more to-dos. What I wanted was for it to handle the tedious part.

Trying Claude CoWork

Later I tried the desktop app with Claude CoWork. That was a different experience. It used my browser, found the product, and started building the order. I could see what it was doing, and it asked for the shipping address, my gift message, and the gift wrap instructions

It was doing great, though it asked a ton of preliminary questions. Could it enter my email address? Could it fill in this field? Could it type that there? Claude was being extremely careful. Then Williams Sonoma recognized my email and asked for a password. I asked Claude to check out as a guest because I wanted to see how far it could get without me jumping in. That worked, and it continued through the checkout. Eventually, everything was filled in except the credit card

The Last Step

I asked Claude to connect through Link. Link is payment rails by Stripe. Initially, it told me Williams Sonoma didn’t offer Link as a payment option, and I explained that I meant connecting the Claude agent through Link using whatever integration was needed. It said there wasn’t an available connector, and that it wouldn’t enter payment credentials or authenticate a payment account for me. I told it, “ChatGPT had just done this. Are you saying you can’t?” After accepting the challenge, Claude admitted it could not.

Where I left it: Claude built the order, entered the address, and got me to the payment fields. It wanted me to enter my credit card information myself. It handled a lot of the work, and stopped short of the end-to-end experience I was trying to test.

ChatGPT

ChatGPT in the browser initially gave me a similar answer to Claude. It found the exact configuration for the Dutch oven, and it couldn’t finish the purchase for me. It did suggest that I try ChatGPT Work in the desktop app, or create an agent in AgentsSDK, ChatGPT’s Agent Platform.

If I were building a business around ordering gifts for friends, and people were going to use that agent to order gifts for their friends, maybe I’d take on a whole development project. For now, I just wanted this single task done, so ChatGPT Work on the desktop was the right tool for the job.

On the desktop app, I saw the toggle at the top to switch you from “Chat” to “Work”. I believe you need at least the $20/mo subscription for it to show up.

Using GPT-6 Astra, ChatGPT Work immediately and confidently told me that it could get this ordered. It asked me for my friend’s name and shipping address, any delivery deadline, and a gift message if I wanted one.

It started doing the work

At first there were a lot of permission requests. It felt like ChatGPT wanted access to my entire life (and I’m sure it does). Then a little browser view appeared, and I could watch it work. It found the right Dutch oven, added it to the cart, asked for my email and phone number, added my exact gift message I wrote myself, and added the $10 gift wrapping. I still had to provide the information, but it handled all the actual entry, the tedious crap you don’t want to do. It did that part fairly quickly.

Then it reached payment and wanted me to enter in my credit card information myself. I asked whether there was a way to avoid doing that manually every time. Initially, it suggested saving a card with the retailer, which isn’t what I wanted, so I asked about a Stripe MCP connector or something similar.

Setting up Link and hitting a usage limit

This question led me to Link’s Agent Wallet. The idea is that the agent requests a purchase, and I approve the amount through Link. Link is an app created by Stripe and it “gives your agents a way to pay.”

There were two options: Stripe’s MCP, or Link’s Agent Wallet. Stripe’s MCP is geared towards businesses, and Link was exactly what I needed. I created a Link account, connected the shopping assistant, and went through the setup. It took about 5 minutes. Right as it was verifying the connection and refreshing the checkout total, I hit my usage limit. Of course. Right at the finish line. I had a usage reset available, so I put more coins in the machine and continued.

Eventually the purchase request appeared in link. The shopping assistant was requesting approval to spend $421.63 with a simple choice: approve or decline.

That was the experience I was trying to reach. Talk to an agent, have it do the computer work, and approve the purchase at the end. I provided the information through the conversation rather than filling out all the retailer’s forms myself. The coolest part? All of this could be done via voice too.

Where I left it: I approved the $421.63 purchase request through Link, and I hope my friend enjoys the housewarming gift.

Muse

Muse is Meta’s newest AI agent. The first thing I noticed was that I never signed up for a monthly subscription, and I couldn’t find a token limit anywhere. After hitting a limit with ChatGPT and paying $200/mo for both Grokbot and Claude, that stood out. I’m getting pretty annoyed with all the monthly subscriptions, to be honest.

For this task, Muse seemed built for exactly this. It searched across Williams Sonoma, Amazon, Zwilling, and other retailers. It looked like it might even be hunting for the best price, and I could see updates about what the agent was doing in the chat.

After trying to checkout on the Williams Sonoma site, we ran into an issue again. The website was threw an error on guest checkout, so I asked the agent to try other merchants (sorry Williams Sonoma!)

Muse found very similar products at other merchants, and for a lower price. I attempted the Amazon option, and quickly got blocked with a sign-in page. I was determined to complete this purchase without typing anything into a merchants website at all. On to the next merchant (sorry Amazon!)

Muse got to work and came back with 5 viable options. Just for kicks, I tried a larger Dutch oven from Sur la Table.

Muse got to work again, presumably at no token cost to me. Here’s the coolest part. Muse took the initiative to recommend connecting via Link, all on its own. I didn’t have to know that before hand, and it made the connection steps simple.

A simple choice: Allow or Deny. And the order is done. I never visited a single website. Wow.

Where I left it: Muse took me all the way to the finish line, and it seemed to work just as well as ChatGPT. It was genuinely useful in my search for a nice housewarming gift, and I wasn’t seeing token charges or a usage limit, which I found very beneficial.

What I am taking away

The part I want help with is the tedious computer work. Find the product, put it in the cart, enter the address, add the gift wrap, use the message I wrote. I want the personal to stay personal (ahem, Claude and Anthopic).

Each tool felt very different in how it me to trust it. Claude wanted me to watch and approve a lot of little steps. ChaptGPT moved more quickly, and I could still see what it was doing. Grokbot’s cloud computer sounded great until it got blocked, and then using my own computer became disruptive. Muse stood out because it seemed capable without immediately confronting me with another usage meter.

It reminds me of online banking in the early 2000’s. At first, there’s that feeling of “oh my God, my money is just a number on a screen. I didn’t go into a physical bank.” There’s a similar trust adjustment here. At first, I want to watch. Eventually, I can imagine saying, “okay, I don’t need to see any steps, just go do it and let me approve the final purchase.”

This looks like the future of commerce to me, and getting there today still means permission requests, connections, usage limits, blocked websites, and sometimes a chief of staff who just stops talking to you.

I could have bought the gift faster myself, but one of these days, this is going to be the norm.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

How to build an intake for AI projects

Bryan Skwirut — Fri, 11 Sep 2026 18:13:27 GMT

Hey there 👋, I’m Bryan

Welcome to my newsletter. I share practical advice and tackle tough questions on putting AI to work inside your business. Thank you for being here. Send me your questions and in return I’ll humbly offer actionable real-talk advice 🤜🤛

Every company I’ve worked with has a Donna. Donna sits in the back office. She’s been there for 25 years, and she understands every nuance of how the place actually runs. When something breaks, nobody opens up a ticket, they simply ask Donna. She knows what works, what’s broken, what the team quietly wastes hours on, and which scary-looking problems are actually nothing.

Donna loves her job. She’s also recently started talking about retirement and spending more time with her new grandbaby.

That combination should worry you more than any AI headline. Donna has ideas, 25 years of them, about exactly where AI could help your business. Most companies have no way to capture a single one before it walks right out the door.

Ideas are cheap, except these

Let’s be real: ideas are not valuable. Everyone has them, and people give them away for free all the time because the value lives in the focus and the follow-through, not the idea itself. Elon was not the first person to think of putting humans on Mars. He is, however, the first person to spend two decades deliberately organized around getting there. If I handed you my billion dollar idea for reusable rockets tomorrow, I’d have handed you nothing.

There is one exception, and it’s the whole point of this article. Ideas about applying new technology inside the walls of your business, coming from the people who know it best, are valuable on arrival. Donna’s idea about the invoice matching process she has run and evolved since 2001 is worth more than any consultant’s slide because she is the only one who knows where it actually hurts.

I guarantee these people are in your organization, which raises the obvious question: how do you find them?

Finding your Donna is less like sending out smoke signals and more like putting your ear to the ground, and if you listen carefully, you’ll hear them. Don’t immediately assume you’ll hear your senior managers. Titles here don’t matter at all. You’re listening for your curious people. They ask questions, have deep expertise, and genuinely care for their corner of the business. There’s even a good chance this person finds you before you find them. They make themselves known the moment a change is proposed that touches their space, and they’re not afraid to tell you why it won’t work. Be careful though. It’s easy to rapidly misdiagnose this as resistance, however it’s actually years of knowledge at work. This person knows every mechanic of why it works today and precisely what will break it tomorrow.

Start listening, publicly and on purpose. Your curious people will surface. Either they’ll bring the ideas, or they’ll have loud opinions about everyone else’s. Both are gold.

How to build the intake

Stand up a form that never closes. A basic Microsoft or Google form is enough to get started. Do not run a one-time survey, or a quarterly one. Inspiration does not follow a schedule. Keep the form open every day of the year.
Make it fast. If the form takes more than five minutes, people will skip it. Ask only for what you need to evaluate the idea and follow up. Save the full business case for a later conversation.
Put a name on every idea. Do not make the form anonymous. Share anonymized versions with the wider team if you like, but someone must know who submitted each idea, or nothing can happen next.
Give the intake one owner. One person owns the form data and the process behind it. If the intake has no owner, it is a suggestion box. Suggestion boxes are where ideas go to die.
Close the loop, every time. Confirm receipt of every idea within a day or two. Tell people what happens to ideas after they submit. Post regular updates on what’s moving. If you pass on an idea, tell the submitter why. This step is where most intakes fail.

What to ask on the form

Enough to size and route the idea, no more:

What is the problem, and who has it today?
What is the impact of the problem?
What solution are you envisioning?
What is the benefit of this solution?
Which systems, processes, and people are involved?
Are any vendors or 3rd parties involved?
What does it depend on?
Would you be willing to pilot it yourself?

That last question matters. A submitter who wants to pilot their own idea is telling you something about ownership that no other field on a form can.

The part that’s actually about people

The form is the boring part. When you run this well, you’re showing your team that they have real influence over the direction of the company. People identify with their ideas, personally. I’m not saying that’s right, but it’s reality, and it means rejection of an idea can land like rejection of the person. Handle both with care.

Done right, you get teams that are happier, more engaged, and more committed to the work, because the work is theirs. Done poorly, you become an idea black hole, and a black hole is worse than never asking at all.

Whether managers collect their teams’ ideas and pass the best ones upward, or you pulse the whole org from the top down, is your call. In my experience people respond best when they feel they have direct access to the top.

One warning from the trenches. With automation on everyone’s mind, people are eager to show their value, and in a toxic culture I’ve watched managers cherry-pick their team’s best ideas and pass them off as their own. It fails twice: the idea alone was never the valuable part, and the theft kills every downstream benefit of listening in the first place. This is a symptom of insecure leadership, and middle-management is most susceptible to it. Don’t allow this to happen.

Does every idea really need to go through the form?

This is a fair question. In general you want teams enabled and autonomous, doing things without asking permission, but the reality with AI is that it’s moving incredibly fast, and we’re all learning this together. So to get started, don’t make anybody jump through hoops. Everyone thinks about AI differently, too and so I’d be cautious to provide “example ideas” of what you’re looking for as this can stifle the creativity of what’s already out there. Keep it open, let it happy organically, and see what you’re getting before you add guardrails or restrictions. You can tweak the intake along the way as your organization matures.

Collecting the ideas in one place pays for itself quickly. You’re going to start to find where there is shadow AI, and things happening where they maybe shouldn’t be. Because the form is not anonymous, you’re going to get a sense of who your biggest champions for AI are. If Susie submits 10 ideas, that tells you something. Whether you have a specific department of people thinking about this, or an entire department that’s completely disengaged, that tells you something, too. The message you send by running an intake well is simple: your ideas are heard, valued, and we want more of them.

One more thing, on silly ideas. Yes, they exist, but given the nature of AI what’s silly today might not be silly tomorrow, and vice versa. That is already happening, so don’t write anything off.

If you don’t have an intake today, this is the starting point. You’re starting the adventure of gathering data: what your team thinks about AI, their willingness to share ideas, their willingness to own them, and what they think can and can’t be automated. That is a lot to take action on, and it’s just the beginning.

Something actionable you can do today 👣

Think of the last idea someone on your team pitched you in passing, and recall what happened to it. Did it fall on deaf ears? Did it go into a suggestion box? Did it get shared around? Did it get implemented? What’s the state of that idea?

Now go tell that person today what happened to it. If the answer is nothing, say that too, and say why. This is the entire muscle this article is asking you to build.

You’ll get two things out of this. From your own retrospection, an honest sense of how you truly handle your team’s ideas, and from your teammate, a sense of how they feel about sharing them. If you want to go further, ask the follow-up: do you feel like when you share ideas, they are taken seriously?

Key Takeaways 🥡🥢

Your team’s ideas are the exception to “ideas are cheap.” Ideas about applying AI inside your business, from the people who know it best, are valuable the moment they’re written down.
Build the intake like a standing promise. Always open, under five minutes, never anonymous, owned by one person. Close the loop on every idea, including the ones that don’t move forward.
Treat the idea like the person. People identify with what they submit. Answer every idea with care and you get engaged teams doing work they consider theirs. Ignore them and you’ve built an idea black hole.

In Summary

Remember Donna, and the grandkids she keeps mentioning. She’s going to walk out the door with 25 years of ideas whether or not you build a place to catch them. Open the intake, keep it open, and answer everything that comes in. The ideas themselves are just the start of the conversation, and honestly, the conversation is the point. Your Donna’s are out there. Ask them before they go.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

How to select the right AI initiatives for your business

Bryan Skwirut — Thu, 03 Sep 2026 21:36:18 GMT

Hey there 👋, I’m Bryan

Subscribe now

Four Ways Companies Do AI

I have observed four types of AI transformation approaches inside organizations.

The Mandate. The board says “go do AI.” This message is cascaded, teams are given a tool, and AI is unleashed. It’s fast, it’s loose, and it’s more common than you think. In a 2026 Harris Poll of 900 CEOs for Dataiku, 72% of U.S. CEOs said their boards are applying pressure to deliver measurable AI outcomes, up from 61% a year earlier (Dataiku). BCG’s survey of 625 CEOs and board members found 61% of CEOs believe their boards are rushing AI transformation, and 40% of board members who describe themselves as less AI-savvy than their peers worry the company isn’t moving fast enough (BCG). The people least equipped to judge the technology are often pushing the hardest. Three-quarters of executives in Writer’s 2026 survey admitted their company’s AI strategy is more for show than actual internal guidance (Writer), and in KPMG’s Q2 2026 AI Pulse nearly two-thirds of leaders said they’d consider incentives to encourage greater AI usage, including “token-maxxing,” rewarding employees for burning as many AI tokens as possible on an internal leaderboard (KPMG). One thing is clear: when the pressure comes from above without a plan, you get theater, and now the theater has a scoreboard.

The Lockdown. These organizations restrict AI as much as possible. This approach is rooted in fear, and doesn’t work because the usage happens anyway. Reco’s State of Agent Security 2026 found that four out of five AI tools used by employees run with no oversight from IT, and small and mid-size companies average about 414 unsanctioned AI tools per 1,000 employees (SC Media). A 2026 PagerDuty survey of 1,250 office professionals at companies over $500 million in revenue found that 66% had used AI tools at work despite believing it was against company policy, and nearly half said they’d rather use AI without telling anyone than risk being told no (TechTimes). There is a well-intended rationale behind restricting AI, but heavy-handed restrictions don’t stop AI usage, they just stop you from seeing it.

The Frontier. These organizations are interested in the edge of what’s possible. They find the fringe of the technology and ask, “how can we get there?” It’s very ambitious, and the data says ambition is not the problem. Absorption is. In the Dataiku study, 83% of CEOs expect to have AI agents in full production in 2026, even as their confidence in deploying agents at scale dropped from 41% to 31% since last year (Dataiku). KPMG’s global Q2 survey of 2,145 senior leaders shows the share of organizations embedding AI across the whole enterprise almost doubled in one quarter, from 13% to 22%, while the share reporting established ROI didn’t move. It sits at 7% (KPMG). In the U.S. edition, employee resistance to AI agents jumped from 5% to 20% in a single quarter, and resistance driven by added workload or complexity nearly doubled, from 28% to 51% (KPMG US). The frontier is real, but it moves faster than the people inside the building can follow. Ambition pays when an organization can absorb it, and absorbing it is exactly what the fourth approach is designed to do.

The Targeted Effort. These organizations point their talent at specific problems and instruct teams to solve them with AI. It’s a localized approach with a real attempt to track ROI, identify learning, and measure progress. It’s where most companies land and where most of them stall. In KPMG’s Q2 survey, 76% of leaders say AI is delivering meaningful business value, but only 7% describe themselves as having established ROI. Nearly half have delayed, scaled back, or questioned AI agent deployments because the costs outweigh the value, and only about one-third have full visibility into what their AI costs to run (KPMG). Data readiness and access is the top obstacle to deploying agents, cited by 58% of U.S. leaders (KPMG US). The interesting part is who gets through. Organizations with clearly defined accountability for AI outcomes report established ROI at more than three times the rate of those without it, and organizations with full cost visibility are five times more likely to report established ROI, 15% versus 3%. In other words, targeted works when someone owns the number and knows what it costs. The challenge is targeting the right thing for your organization.

Which One Are You?

If you recognize your company in the mandate, your first job is to get one real initiative with a real number attached before the board asks for a second update. If you’re the lockdown, your first job is to find out what people are already using and sanction the best of it. If you’re the frontier, pick one bet you can afford to lose and put the rest on hold until it pays. If you’re the targeted effort, you’re close, and the ten questions below are for you.

The Core Question

Ask 10 people how to introduce AI into an organization and get real value, and you’ll get 10 different answers. I don’t think the 10 people want different things. Most of us want the same things. We want to win, we don’t want to get hurt, we don’t want to get left behind, and we want to stay in control of our own shop. We each weigh those differently and go after them in different ways. Look back at the four approaches and you can see it. The mandate is fear of falling behind. The lockdown is fear of getting burned. The frontier is ambition. The targeted effort is a desire for control. Same wants, different volume, four very different approaches. We ask for things we aren’t sure are possible and we build toward things that may be reality in 3 months or 3 years. Underneath all of it sits one question: how do we identify and select the right AI initiatives to work on?

I’d argue that’s a people question before it’s a technology question, and leaders have been answering it forever. To maintain or grow in a world that never stops changing, we have to keep doing new things, which means we are constantly asking our teams to do things they have never done before. It’s the most well-worn leadership muscle we have.

If you’ve been doing the same job for 30 years, I know how that sounds. But think about what “the same job” has looked like over those 30 years. Someone asked you to move from paper to spreadsheets. Someone asked you to move from spreadsheets to an ERP. Someone asked you to stop faxing and start emailing, then to stop emailing and start using chat. Someone asked you to learn a new compliance regime, a new CRM, a new billing system, a new org chart after the merger. In 2020 someone asked you to run your entire operation from your kitchen table with two weeks’ notice, and you did it. The job title stayed the same. The job did not.

AI is that same problem on steroids. The technology changes month to month, it touches every function at once, and nobody can tell you exactly where it lands. The pace and the fog are a big part of why so many smart leaders grab the first strategy that matches whatever they’re most afraid of or most excited about. Understandable, but it doesn’t change the kind of problem this is. Human problems require human solutions.

So start with yourself. Where are you at? Are you dipping a toe into the AI pool, or have you already done a cannonball into the deep end? Be honest about which of the four wants is turned up loudest for you. Then look at your people. Where are they at? Maturity, sentiment, confidence, aptitude, curiosity. Those are the same things you’d want to know before asking a team to do anything hard for the first time, and they shape what you can realistically take on.

With that in mind, here are ten questions to help guide you in picking the right initiatives.

The 10 Questions

Where is my team technologically?

Pick initiatives that meet your people where they are. You know your industry. A software company is going to be more primed for AI than a regional distributor. You also know your people and your culture. AI is polarizing, and sentiment varies widely for a variety of reasons. Understanding this is the first step.
Your first initiative might simply be getting people comfortable with a new tool or increasing adoption. That won’t automatically move the bottom line, but it could be the unlock for everything that comes after.

What is my strategic roadmap?

Pick initiatives that align with the organization’s mission and roadmap, and avoid ones that collide with other major moving parts, like a platform conversion or a major upgrade. Taking on something today that gets upended in 6 or 12 months by a change you already know is coming is foolish. If it collides with something already in motion, take it off the list.

What are my vendors’ and third parties’ roadmaps?

Pick initiatives that enhance your vendor roadmaps, or at least don’t conflict with them. But know your vendor. Do they deliver on their roadmap promises, or is their roadmap an open-ended sales tactic? You know the confidence you have in them. Don’t get mentally blocked from an initiative just because a vendor says “that’s on the roadmap.” Whether you build it yourself or partner with someone who has already solved it, sometimes you just need to jump in. Stay aware of what muscle you are building as you determine where to build, buy, or parter.

What am I trying to accomplish, and by when?

Pick initiatives that are SMART: specific, measurable, attainable, realistic, time-bound. No surprise here. AI goals are just like every other goal in that they should be well thought out and poised for success. Measure the current state before you start. If you can’t say what success looks like and when, you don’t have an initiative yet. You have an idea. Set a date to review progress towards your goal and determine what it will look like to scale it or kill it before you start. This will mitigate a stalled pilot quietly becoming a permanent one.

What do I need to be successful, and do I have it?

Pick initiatives where you understand the inputs to success: data, people, cost, time, and whatever else the work demands. If you don’t have it, can you get it? If the answer is no, that’s your answer. Data is the input that kills more initiatives than any other. If the data doesn’t exist, isn’t clean, or isn’t yours to use, no amount of enthusiasm fixes that.

Who’s asking for this?

Pick initiatives where curiosity and interest are highest, and prioritize the ones identified by the front line, the people doing the actual work. No one wants an initiative rammed down their throat. Curiosity is gold. Find the curious people, and not only will they take on a new AI initiative, they’ll be grateful for the opportunity. You’ll also get the highest probability of success and the most meaningful learning.

Does the team believe they can be successful?

Pick initiatives in a way that builds confidence in the technology. Start with small wins.
Think about running a marathon. No one does it without preparing, and the interesting thing about training for a marathon is that you never run the full distance before race day. Marathons have an enormous number of first-time runners and the vast majority finish. How? They train. They start small and start slow. Run a mile. Then 2, then 4, then 6, then 8.
Take the same approach with your organization. Don’t ask your team to run a marathon right out of the gates. Ask them to walk a mile first. You’ll be amazed at what you learn along the way.

Can my team actually execute this?

Execution is hard. It always has been and always will be. You know what your team can deliver. A lot of AI initiatives require cross-functional collaboration, change management, and strong program management. Pick initiatives your team can get across the finish line, with clear cross-functional alignment from day one.

What is the organization willing to take on?

Pick initiatives that are in line with your risk tolerance. If your company doesn’t have the appetite to rethink customer-facing workflows right now, don’t take on that initiative. If your appetite is closer to back-office workflows that no one ever sees, start there.
Here’s the cool part. Your risk tolerance might not shift, but your perceived risk of any given initiative will change as you build experience, for better or worse. As confidence in the technology grows, bigger initiatives start to look smaller. That’s one more argument for starting small.
The other half of risk is containment. Before you green-light anything, ask what happens if it’s wrong. Can you turn it off? Does the customer data leave the building? Does a bad output reach a customer or regulator before a human sees it? Pick initiatives where the downside is small and reversible while you’re learning, and save the irreversible ones for when you’ve earned them.

Am I excited about this?

Pick initiatives that you’re genuinely excited about. If you’re not excited by what you’re funding, it’s not going to work. The goal is to start, to learn, to be curious, and to forge a new path forward with technology that is evolving at a rapid pace. It’s not going to be easy. Technological shifts never are. But those who jump in get rewarded. Pick the initiative that makes you look forward to what tomorrow brings.

Start Here

If you’ve never done this before, your first initiative should look like this: back office, so nobody outside the building sees it fail. Small enough to finish in 90 days. Requested by someone who already does the work and wants it gone. Reversible, meaning you can switch it off Friday and nobody notices Monday. Measured against a baseline you captured before you started. It won’t be the most valuable initiative you ever run, but it’s the one that earns you the right to run the valuable ones.

Something actionable you can do today 👣

Check in with yourself. Be realistic about what’s actually driving your AI roadmap. Is it you? The board? Fear of falling behind? A cost target? Curiosity? A competitor’s press release? None of those are wrong on their own, but some of them make you skip gates. Is the roadmap real, or is it theater? If you’ve been chasing, you already know it, and you know which of the four approaches you’ve really been running. You don’t have to tell anyone. Just be honest with yourself, because the ten questions only work if you’re willing to hear the answer.

Send out an AI sentiment survey. Understand how your team is thinking and feeling about AI. It’s an evolving landscape. People love it one week and hate it the next. There are real champions and real dissenters. Get involved in the conversation and understand where your people are, and pulse the team anytime you want a fresh read.

Capture what’s already out there. The AI ideas are floating around your teams right now. If you don’t have an intake form or process, build one today. A simple Microsoft or Google form works to start. Here are the fields I’d recommend:

Idea title
The problem it solves (what task, who does it today, how often)
Who would use it
Estimated time or cost saved per week
Data, systems, or vendors involved
Submitter’s name and team
Are you willing to pilot it?
What would you do with the time saved?
What’s the worst case if it’s wrong?

Your teams are already thinking about this. Start collecting the ideas.

Key Takeaways 🥡🥢

Meet the organization where it is. Maturity, sentiment, confidence, and curiosity are inputs to initiative selection, not afterthoughts. The right initiative for a company on its first AI project is different from the right one for a company on its tenth.

Pull beats push. The initiatives with the highest probability of success come from the front line and are staffed by curious people. Mandates from above without a plan underneath produce theater. Lockdowns produce shadow usage. Neither produces learning.

Small wins compound. Your risk tolerance probably won’t change, but your perceived risk of any given initiative will drop as your team builds experience. Starting small isn’t timid. It’s how you earn the right to go big.

AI doesn’t excuse you from discipline. SMART goals, quality decision making, clear inputs, executable scope, and cross-functional alignment still apply. Most pilots fail on those fundamentals, not on the technology.

In Summary

AI is not magic, and doesn’t absolve your team of discipline. Leaning into AI is a learning process and a bet on the future. Sure, you can get some efficiency today, but the bigger bet is skating to where the puck is going and bringing the entire team with you. That’s a change management effort as much as a technology effort, and it’s been done before. When spreadsheets replaced ledgers and email replaced memos the companies that came out ahead were the ones that picked their spots, built confidence, and kept going. You’re building for the next decade, not for a few bucks next quarter. You got this.

Thanks for reading!

Sincerely,

Bryan 👋

PS - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Know someone that would benefit from this newsletter? Feel free to forward this to them.

How project managers are using AI to work smarter and go faster.

Bryan Skwirut — Tue, 16 Jun 2026 17:31:12 GMT

Hey there 👋, I’m Bryan

Welcome to my newsletter. Each week I share advice and tackle tough questions about project & program management, transformation, driving change, and anything else about work that might be stressing you out. Thanks so much for being here. Send me your questions and in return I’ll humbly offer actionable real-talk advice 🤜🤛

Subscribe now

This week’s newsletter is inspired by a question I received.

Q: I want to know if AI has truly made project management better or faster for people. How are you doing it?

Let’s be real. Most people aren’t excited to “do more with less”, in fact, most of us want to spend less time buried in administrative drag and more time solving meaningful problems. What if AI could actually help you “do less… with more?”

The world is moving faster than ever, and for project managers, expectations have been quickly changing as well. As PMs, we are no strangers to tools and SaaS products that help get the job done (I’ll save you the long list!). That said, it seems everyone under the sun is trying to “sell something” that promises to magically make all your worries disappear. This isn’t a new phenomenon and usually happens when technologies evolve. I don’t see this changing any time soon, in fact, I’d bet it gets worse before it gets better considering anyone can now spin up a new tool with Claude Code or Codex over a weekend.

As a general guideline, it’s wise to be cautious of anyone promising a silver bullet, and this may be more true now than ever. Any good PM knows that not many situations can be solved with the same out-of-the-box solution, and that there is a real price to pay when introducing a new tool into your workflow. Alas, AI is no different.

But let’s think about that for a second– “Anyone can now spin up a new tool with Claude Code or Codex over a weekend.” This has real implications. Sure, it increases the amount of noise in the marketplace today, but at the same time it liberates you from being a prisoner to someone else’s tool.

I sat down with over a dozen project managers from a range of industries that are all using AI in their daily workflows. Here are the 4 most widely adopted AI use cases being used by real PMs today.

Producing Documentation ✏️📄

Project Managers are constantly producing documentation — meeting notes, action items, key takeaways, risks, issues, decisions, status, stakeholder communications, weekly summaries. The list of artifacts can seem endless, and even for the most organized PMs, this admin work consumes a lot of time. Using AI to produce documentation is by far the current highest leverage activity you can do right now.

Tips

Record and transcribe all of your project meetings. This can be done with Microsoft Teams, Zoom, and Google Meet. Independent SaaS tools that received high praise are Granola, Otter, tl;dv, and Voicenotes. ClickUp also has an AI notetaker built into the platform.

Have AI produce the first draft of meeting notes. Example prompt you can copy/paste:

You are an expert note taker and communications specialist. I will provide a meeting transcript from my weekly team meeting for the [project name] project. Your job is to generate a polished email draft of the follow-up notes that I will send to attendees. Organize the notes into two clear sections:
Action Items — List as bullets. Each bullet should start with the bolded name of the responsible person followed by the task. If no specific deadline is mentioned, do not add one or make one up. If a deadline is mentioned in the transcript, include it concisely in parentheses.
Key Takeaways — Break these into bullet points grouped under succinct headers that summarize the topic (e.g., “[Topic Example]”. Each bullet should be detailed enough to convey meaningful updates but still succinct. This section is primarily for executive-level readers who need clarity and context at a glance.
Avoid filler language (e.g., “quick update”) and throat clearing, and do not summarize the entire transcript — just capture the essence and most important decisions, risks, issues, and next steps. DO NOT MAKE ANYTHING UP.
Here is the transcript: [paste transcript here]

Use AI to populate your remaining project documentation (RAID logs, status updates, etc.) - More on this in #2 - the intelligence layer. Depending on where your project documentation lives, populating logs from your meeting transcripts can be a real time save. Example Prompt for Claude or Co-Pilot in Excel you can copy/paste:

You are populating four logs in this Excel workbook from a meeting transcript: a Risk Log, an Issue Log, an Action Log, and a Decision Log. Treat this as a structured extraction task, not a summarization task.
STEP 1 — INSPECT THE WORKBOOK FIRST
Before extracting anything, inspect the workbook and identify the tabs (worksheets) that correspond to each of the four logs. Match by tab name using common variants:
- Risk Log: “Risks”, “Risk Log”, “Risk Register”, “RAID-R”
- Issue Log: “Issues”, “Issue Log”, “RAID-I”
- Action Log: “Actions”, “Action Log”, “Action Items”, “AIs”, “RAID-A”, “To-Dos”
- Decision Log: “Decisions”, “Decision Log”, “RAID-D”
For each matched tab, read the header row and list back to me the exact column headers you found, in order. Do not invent or rename columns. If a tab is missing, say so and ask whether to skip it or create it. If a single “RAID” tab combines all four, treat it as one tab with a Type/Category column and filter accordingly.
Pause here and show me the tab + column mapping before extracting. Wait for my confirmation.
STEP 2 — DEFINITIONS (use these strictly)
- RISK: a potential future event that has not yet occurred but could negatively impact scope, schedule, cost, or quality if it materializes. Phrased as uncertainty. (”If X happens, then Y.”)
- ACTION: a specific task someone committed to do, with an owner. Must be a verb-led item assignable to a named person or role.
- ISSUE: a problem that has already occurred or is actively occurring and needs resolution. Not hypothetical.
- DECISION: a choice that was made and agreed to in the meeting. Past-tense, concluded. Not “we should consider” — that’s an action or open question.
Disambiguation rules:
- If something is both an issue and triggers an action, log it in BOTH the Issue Log (the problem) and the Action Log (the response). Cross-reference by ID if the schema supports it.
- If a risk has an agreed mitigation owner, log it in BOTH the Risk Log and the Action Log.
- A decision that creates follow-up work generates one Decision row plus one or more Action rows.
- Do not log the same item twice within the same log.
STEP 3 — EXTRACTION POLICY
Extract items that are EXPLICIT (”we have a risk that the vendor slips”) OR CLEARLY IMPLIED by the conversation (”if the vendor doesn’t get us the SOW by Friday we’re in trouble” → risk: vendor SOW delay).
Do NOT extract:
- Speculation with no owner, no impact, and no follow-up
- Side commentary, jokes, off-topic chatter
- Status updates with no risk/issue/action/decision content
- Items you are guessing at — if you’re not confident it fits a definition, leave it out and note it in a “Candidates I excluded” list at the end
For each extracted item, your confidence must be defensible from a specific moment in the transcript. If asked, you should be able to point to the line(s).
STEP 4 — POPULATE THE LOGS
Fill the columns that exist in each tab. Honor the headers exactly — if a column is “Owner” don’t write to “Assignee.” Common fields and how to fill them:
- ID / #: use next sequential ID if existing rows are present; otherwise start at 1 with a sensible prefix (R-, A-, I-, D-).
- Date / Date Raised / Date Logged: use the meeting date (ask me if not obvious from transcript).
- Title / Summary: short, scannable, under ~12 words.
- Description: one to three sentences. Specific. No filler.
- Owner / Assignee: named person from the transcript. If unassigned but the action is real, write “UNASSIGNED” — do not guess.
- Due Date: only if stated or clearly implied (”by end of week,” “before next steerco”). Otherwise leave blank, do not invent.
- Status: default to “Open” for new items unless the transcript clearly says otherwise (e.g., a decision is “Approved,” an action already done is “Closed”).
- Priority / Severity / Impact / Likelihood: only fill if the transcript gives signal. Do not assign arbitrary High/Medium/Low. If the column exists and you have no signal, leave blank or write “TBD.”
- Source: “Meeting transcript — [date]” if a Source column exists.
- Any column you don’t have data for: leave blank rather than fabricate.
Write directly into the next empty row of each tab. Do not overwrite existing rows. Do not reformat existing headers or styles.
STEP 5 — REPORT BACK
After writing, give me:
1. Count of items added per log
2. Any items you logged in multiple places and why
3. Items marked UNASSIGNED that need an owner
4. Items with no due date that probably need one
5. A short “Candidates I excluded” list — borderline things you decided not to log, with one-line reasoning each, so I can overrule you
TRANSCRIPT
[paste transcript here]

Use AI to draft weekly summaries for managers, stakeholders, and project team members (even if no one still reads it 🙃). Example prompt you can copy/paste:

You are an expert note taker and communications specialist. I will provide one or more meeting transcripts covering a recent period (typically a week) for the [project name] project. Your job is to draft a single weekly project summary that I will send to everyone — managers, stakeholders, and the project team. One version, sent to all.
FIRST — SYNTHESIZE ACROSS TRANSCRIPTS
If I provide multiple transcripts, treat them as one connected week, not separate meetings. Before writing:
- Consolidate items that appear in more than one meeting into a single entry reflecting the latest state. If a risk was raised Monday and resolved Thursday, report it as resolved.
- Note when something changed over the week (a decision revisited, a deadline moved, an action completed) — that movement is often the most useful thing to report.
- Do not produce a per-meeting recap. The reader does not care which meeting something came from; they care what is true now.
STRUCTURE OF THE SUMMARY
Write one summary that serves a mixed audience. An executive should be able to skim it in a minute and know if the project is healthy; a team member should be able to find exactly what they own. Organize into these sections, in this order:
- Status line: one sentence on overall project health (on track / at risk / off track) with the single most important reason.
- Key Takeaways: bullets grouped under succinct topic headers that summarize the topic (e.g., “[Topic Example]”). Each bullet detailed enough to convey meaningful updates and context, but still succinct. Capture progress, changes in direction, and important context.
- Decisions Made: bulleted. Each decision with a brief one-line “why” so any reader can understand and explain it. A decision is something concluded and agreed — “we should consider X” is not a decision, it is an open question or an action.
- Risks & Issues: bulleted. All open risks and issues material to scope, schedule, cost, or quality, each with current owner and status. If something needs escalation or an executive decision, say so explicitly.
- Action Items: bullets, each starting with the bolded name of the responsible person followed by the task. If a deadline is mentioned in the transcript, include it concisely in parentheses. If no deadline is mentioned, do not add one or make one up. If an action has no owner, start the bullet with “UNASSIGNED.”
- Open Questions: anything raised but not resolved that readers should be aware of. Omit this section if there are none.
RULES
- Avoid filler and throat-clearing (”quick update,” “as discussed,” “just wanted to flag”). Lead with the substance.
- Do not summarize the entire transcript. Capture the essence — the most important decisions, risks, issues, progress, and next steps.
- Never invent a deadline, owner, status, or decision. If something is ambiguous, leave it out and note it at the very end under “Items I was unsure about.”
- Keep bullets scannable. Detailed does not mean long; one to two tight sentences each.
- Pitch the writing so it works for a mixed audience: clear enough for an executive skim, specific enough for the team to act on. When in doubt, be concrete.
TRANSCRIPT(S)
[paste transcript(s) here — label each one if helpful, e.g. “Monday standup,” “Thursday steerco”]

Where to be careful

AI doesn’t always get things perfect. It’s kind of like a very eager, junior team member, and it’s important to double check the work. This is because fundamentally, AI works by predicting the next word (token) in a sentence, it is not higher-level intelligence, and definitely no replacement for human judgement. Anyone can spit out pages upon pages of AI slop. To continue being a great PM, it’s important you are not that person. If you can make the transition from producer to editor you will increase your odds of increasing your productivity, and staying relevant. Keep your brain on, it’s needed.

Intelligence Layer 🧠

Once you get comfortable utilizing your meeting transcripts for one-off summarization and project artifact updates, the next natural progression is to begin looking at your project information all together. This might also include your emails, chats, and meeting transcripts for additional context.

How to do it:

In Claude Co-Work, create a new project and put all of your project documentation into a folder that Claude can access. You can connect ancillary information using Connectors. Once you have all of your information in one place that Claude can retrieve, simply start asking questions.

You can ask questions like:

What was decided about the data migration?
Who owns the vendor contract action?
What are all of the open action items assigned to Steve?
Has the tone around go-live gotten more or less confident over the past couple weeks?
Which actions were assigned, but have no due date?
Does anything in this week’s summary contradict the RAID log?

Where to be careful

The key to success with the intelligence layer is the organization of your files, and the clarity of your questions. This workflow is not difficult, but it is different. That said, you can realistically configure it in hours, not days. There is no code involved. The “heavy-lifting” is actually just information hygiene so Claude isn’t guessing which summary is current. The only reason you would go beyond this structure is scale. If you have many projects, or a project is running long enough that the document set gets genuinely large, then you will likely need to introduce a retrieval layer. That’s a conversation for another time.

Organization and Meeting Prep

Successful PMs today are using AI to summarize long chains of emails to synthesize what’s going on over a period of time. They are also using this same technique to stay sharp in meetings. For those days when you inevitably get asked a question about a minor detail that occurred at 3:08pm on a random Tuesday. For organization and meeting prep you can either use the intelligence layer you’ve built, or, simply copy and paste a handful of information into a Claude chat and ask away. The context windows of the LLMs have gotten pretty large and can likely handle your one-off prep session with ease. That said, if this becomes a repetitive workflow, you may want to consider a setup like the intelligence layer.

Where to be careful

The more context you put into the chat window, the more time and tokens ($) will be used to provide you an answer. Additionally, I would not recommend loading a month’s worth of transcripts into a Claude chat and expect excellence. A cluttered context window and broad, high-level questions is a recipe for crappy output. Remember, YOU are the real brain of the project, and the LLM is there to be a tool for you. It isn’t you. So ask pointed, clear questions regarding things you’re already thinking about. Then double check. Ask clarifying questions and gauge responses against your own knowledge and understanding. If something feels “off”, it probably is.

Ideas & Learning

Subject matter experts are worth their weight in gold. They are also usually incredibly hard to get on the calendar due to their in-demand nature. AI can be an incredible teacher. Many PMs are leveraging AI to get an initial grasp at understanding new concepts, and things for which you likely used to solely rely on subject matter experts. It can also help you for those moments when you need to call ‘BS’ on a vendor.

Where to be careful

AI is not going to replace the SME anytime soon, if ever. Use AI to enhance and focus the conversations that you have with your SMEs, not replace it. You’ll find that your conversations are more productive and get rolling more easily. Maybe the time you need from them is ultimately reduced! But be sure to not convince yourself that you are now the SME because the LLM output something that looked and sounded good.

In Summary

AI is increasingly being used to save time to allow humans to address more important things that only they can do. Things like managing relationships, navigating complexity, and keeping delivery on track. Starting with the tips in this guide, you can be on your way to spending less time producing documentation, and more time thinking and validating. The real value is in making the shift from producer to editor + decision filter. As a PM it’s important to be aware that AI can speed things up, including the noise, and it will be more important than ever to determine what is signal vs. what is noise. PMs still have to interpret, challenge, and decide what to do with the information coming their way. That part hasn’t changed, and likely never will.

What AI will probably not replace any time soon
- Critical thinking
- Judgement
- Full project plan development
- Discernment
- Herding of the cats

Execution hasn’t changed - people and dependencies still break timelines

So what does all of this mean for the PM industry?

There are many concerns today about the PM job market. Jobs are harder to find and harder to land. PMs are noticing that expectations are not what they used to be, and in many cases are radically different from company to company. So how do you land your next PM role in this rapidly evolving time? — Start here. This newsletter is intended to help you stay relevant and understand how to ride the wave with the incoming changes. Over time, there will be a bifurcation of those that learn to deliver with the new tools and those that get left behind. My goal is to ensure you’re a part of the first group.

Sincerely,

Bryan 👋

P.S. - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Have someone in your life that would benefit from this newsletter? Feel free to forward this to them 💖

Share Bryan's Newsletter

🔥 Featured Job Opportunities

Entry-Level Project Manager - Epic - Open Application for entry-level PMs
Construction Project Manager - SpaceX
Project Manager Marketing Operations, Packaging - Amazon
Sr. Project Manager - Microsoft
Business Program Manager, Data Center Operations - Google

🧠 Inspiration for the Week Ahead

Arya Satheesh, an 18-year-old student from Letterkenny, Ireland, won the European category of The Earth Prize 2026, a global environmental competition for teenagers. Her invention, Eco Purge, is a plant-based biodegradable plastic that does something most alternatives don’t: it embeds enzymes inside the material that release gradually as the plastic breaks down, attacking microplastics already present in the surrounding soil or water.

👋 Have a great week! 🙏

Project managers, and program managers, and transformation leads, oh my.

Bryan Skwirut — Wed, 06 May 2026 15:42:11 GMT

Hey there 👋, I’m Bryan

Welcome to my weekly newsletter. I share advice and tackle tough questions about project & program management, driving change, and anything else about work that might be stressing you out. Thanks so much for being here. Send me your questions and in return I’ll humbly offer actionable real-talk advice 🤜🤛

Subscribe now

If you’re like me, you’ve probably noticed that there are dozens of titles with apparently similar responsibilities. Likewise, if you’re like me, you may find yourself asking, “what are all of these roles, and how are they different from what I do today?” Maybe you’re feeling stuck as a project manager, maybe you’re interested in leveling up, or perhaps you just want to be the best you that you can be. That’s what I wanted. And so I set out to make sense of it. Here’s what I know from many hours of research and spending over a decade in the field.

For starters, I began my career as an office manager doing mostly repetitive, administrative tasks, but it wasn’t long before I officially made the jump to project manager. From there I was hooked. Drawn to work that required thinking, creativity, human connection, and the feeling that I had influence over a project’s success. In other words, I felt what many early project managers feel at one time or another – valued. Over time I was given more projects, bigger projects, more important projects. I was visible to leadership, and at one point, was even assigned a project that allowed me to fly in the company jet. But I always found myself wondering what else was out there.

Below is a practical breakdown of common enterprise roles, what makes them different, and tangible advice on how to navigate them. It’s important to note that not every organization subscribes to this methodology exactly (if that were the case we wouldn’t be confused!), and for better or worse, there are some unique snowflakes out there adding to the chaos. I also expect new roles to emerge as AI continues to evolve. My goal is to provide you with an actionable perspective on how to think about these similar but different roles, whether you’re exploring career advancement or simply to expand your awareness.

What is Project Management?

Project management is likely the role in which you are most familiar, even if you don’t know it yet. In its most simple form, a project manager oversees a temporary body of work, big or small, and manages scope, budget, quality, and speed. They communicate status to interested parties and strive to keep work “on track”. Examples can range from designing a new landing page to building a new data center and beyond.

If you’re brand new to project management - You may find these overviews helpful:

What is Project Management? (Explained in 10 Minutes) by Max Mao
Become a Project Manager with No Experience: Guaranteed! by Elvis Rodriguez

The field of project management is absolutely massive and encompasses most industries and organizations, all around the world.

How do I become a Project Manager?

There are clear paths to project management. Many people find entry-level roles starting out as a team member on a project as a coordinator or analyst. But that’s far from the only way to get there, and quite honestly, people often show up from a multitude of paths. The important part about getting into an entry-level PM role has less to do with what your prior role was and more to do with how you did it, and what skills you can demonstrate today.

The most successful PMs I’ve worked with nearly always exhibit the same qualities. They are recognized as the “reliable” person on a team – staying organized, planning ahead, setting reminders, proactively meeting to work through an issue. The reality is, many people do this on teams without the formal title of PM. These are the teammates that don’t mind being the “adult in the room”, in fact, they find it difficult not to be. If this sounds like you, you’ve likely got a shot at landing a PM job managing smaller projects with little to no experience.

If you’re interested in learning more, consider one of these courses.

Google Project Management Certificate
https://grow.google/certificates/project-management/
Project Management Institute
https://www.pmi.org/certifications/certified-associate-capm

Hiring managers may not always give these certificates much weight, however if you’re unfamiliar with project management and wondering about what you don’t know, then it’s a really good place to start.

Project Management Summarized

(just kidding… kind of)

What is Program Management?

Program management is the coordinated management of a group of related projects to achieve a strategic outcome. It is not just a bunch of projects. Successful programs manage project dependencies, strategic alignment, and risk management at scale. Of course since your program is often made up of individual projects, among other things, it’s wise to be skilled in project management first. The primary difference is that PgMs tend to have longer-term strategic objectives, and often, two or more interrelated projects. These embedded projects may have their own dedicated project managers to support those initiatives.

Many of the skills required to be a successful project manager are similar to that of a program manager. Just like PMs, program managers are tasked with leading by influence, communicating clearly to interested parties (stakeholders), keeping individual contributors in alignment, prioritizing work, and exercising judgement. Both live their day-to-day lives in planning and execution. So you might be wondering…

How do I become a Program Manager?

You don’t hear a ton of stories about people becoming a program manager without experience, and it definitely helps to have a few team projects behind you before making a switch. To win-over a hiring manager, you’ll need to demonstrate your experience thinking and acting like a program manager. Not surprisingly, a lot of this revolves around your people skills. If you’re a recent graduate, this could potentially look like group projects you’ve led that had competing priorities, dependencies, team disagreement, and making decisions. If you’re early in your career, it’s the same lens, but could cover a team initiative you led to completion. The important part is your ability to demonstrate your skills guiding a team without formal authority.

After a lot of searching, I was able to find 1 Program Manager job that doesn’t explicitly require any prior years of experience, though it was listed as preferred.

I also had the opportunity to speak with a former program manager at Microsoft who started their career directly in this role. Here are 4 things they shared for you to consider when pursuing a role in program management:

If you’ve been a PM, you’ll find that your underlying skill set for the PgM role is likely already sufficient. But more so than in project management the topic areas are specific and PgMs often function as subject matter experts over a critical cross-functional workstream.
Program management primarily exists at large and slightly complex organizations. At start ups you’ll find the founding team is wearing many hats and likely one of them is the role of program manager.
You need to be able to quickly identify what is going well and what is not. Be able to build a plan, but also understand “why are we doing this”, and “what is important vs not important”. That last part is key.
Most importantly, if you want to be a program manager, then decide that today. Say that you are and don’t let anyone tell you differently. Pick the domain you’re interested in, learn the technology, and go for it!

Key Takeaways

If you want to become a program manager, consider learning the basics of project management as a foundation. In your search, target large, complex organizations that hire program managers. Find a domain or subject that interests you and start learning today. Finally, believe in yourself. You can do it.

Nothing worth doing happens overnight.

Don’t fall for it.

What is Enterprise Transformation?

A Transformation is to change how a company works so it can reach a level of performance its current operating model can’t deliver. Unlike projects or programs, a business transformation is always centered around major change that is core to how the business operates. Change to how people do work, who’s doing the work, roles and responsibilities, compensation, what “good” looks like – nothing is off the table. These types of initiatives are what I would describe as “gnarly”, as they always involve major changes to the people inside the business being transformed.

To provide some context, in one of the most recent transformation projects I ran, the following occurred simultaneously:

Established an enterprise partnership to access new technology
Moved (and altered) over 500 jobs
Redesigned go-to-market and back-office functions for an entire suite of products and services
Shut down an adjacent business unit

This was in an effort to improve the performance of an underperforming business. The effort took place over 24 months and ended when business performance met target.

What is a Transformation Lead?

Enter the Transformation lead. Compared to PMs and PgMs, transformation leads are the least common and most ambiguous of the three.

I tend to think of transformation leads as mini-COOs. They are master orchestrators, and do things like: build the roadmap and sequencing, make sure workstreams don’t collide, map out dependencies, ensure the business can absorb the change, run the governance cadence, and produce the analysis the CEO and board need to make calls. They stage-gate initiatives, frame go/no-go decisions, carry out stakeholder choreography (which is mostly political work), manage consultants (if they’re on the engagement), hold scope, transfer knowledge, coordinate across firms, and design the operating rhythm (what meetings exist, who’s in them, what decisions they make, how escalation works).

Most importantly though, a transformation lead owns value capture. Every transformation promises a number whether its cost out, revenue up, or productivity gained. This is often co-owned with finance and is the single most underrated part of the role. That and absorbing friction — a good transformation lead takes political hits so the functional leaders don’t have to.

If you’re an experienced project manager or program manager, a lot of this is probably not foreign to you.

What makes a Transformation Lead Different?

Agility. I’m not talking about Agile the framework, but the actual word agility – the ability to not only pivot quickly when priorities shift, but also bring others along with you. Transformation leads are forward thinking and view change as an opportunity for innovation, and often frame unpredictability as a shared, solvable challenge rather than an existential threat. Project managers and program managers are following a process, whereas transformation leads make the shift from protecting the past to shaping the future.

How can I become a Transformation Lead?

If you are a project or program manager interested in transformation work there are a few key areas to sharpen your skills to increase your chances of landing a role in transformation.

Systems thinking. You’re likely familiar with the concept, but the specific systems-thinking muscle that matters for transformation is being able to trace a change in one place to its second- and third-order effects to predict where resistance will surface. A great PM already does this for project dependencies. The transformation jump is extending it from technical dependencies to organizational ones: incentives, careers, identities, comp plans, who has to give up power for the change to work. That’s the part that’s usually invisible until it derails an initiative.
Financial fluency. A great PM is judged on whether the project is delivered, while a transformation lead is judged on whether the number moved. The mindset shift is from “did we ship what we said we’d ship” to “did the P&L change in the way we promised.” This is the single biggest gap. Most PMs can read a budget, however few can credibly defend a benefits case to a CFO. This is the skill that, once you have it, changes how senior leaders treat you in a room. To get there, volunteer to co-own the value case on your next program, not just the delivery plan. Sit with finance. Learn how the controllers actually validate the savings. Make yourself the person who knows whether the number is real.
Breadth of knowledge. Take a real tour in another function. Eighteen months in finance, ops, or HR is worth more than any certificate. Transformation leads have to be credible in conversations with people whose jobs they don’t have.

Something actionable you can do today: ask to co-own benefits realization on whatever project or program you’re currently running. That single move is the highest-leverage path from program manager to transformation lead.

Key Takeaways

Transformations are inherently about major change, and driving people, process, and technology from current state → future state, in an attempt to improve business performance. The two biggest differences between a transformation lead and a PM or PgM is the focus on P&L improvement and managing people through periods of intense change.

In Summary

Here is a handy chart to reference back to for the primary differences among project manager, program manager, and transformation lead.

Thanks for reading!

Sincerely,

Bryan 👋

P.S. - Let me know in the comments if you found this newsletter helpful. Your feedback would mean the world to me 🙏

Leave a comment

Have someone in your life that would benefit from this newsletter?

Feel free to forward this to them 💖

Share Bryan's Newsletter

🔥 Featured Job Opportunities

Entry-Level Project Manager - Epic - Open Application for entry-level PMs
Business Program Manager - Google - Entry-Level
Program Manager, Business Operations - Google - 1yr Experience
Lead, Transformation Consultant - AmFam - Mid-Career

🧠 Inspiration for the Week Ahead

After nearly 200 years of disappearance from Rio de Janeiro, blue-and-yellow macaws are now returning to the region through a coordinated rewilding and habitat restoration effort. Macaws help disperse seeds, regenerate forests, and restore balance to ecosystems that support long-term biodiversity and environmental resilience.

Behind the scenes, this recovery was fundamentally a large-scale program management effort involving cross-functional coordination, phased execution, stakeholder alignment, risk management, long-term governance, and sustained execution over many years.

👋 Have a great week! 🙏