We rolled out AI tools. Why isn't anything changing?

If you rolled out AI tools and nothing is changing, it's because nobody has actually changed how the work in your business gets done day to day. We call this phenomenon “AI theater,” and right now most of the companies I talk to are doing a community play version of it.

Let me guess. You’ve given everyone on your team (or at least on your “AI Committee”) a subscription to Claude, ChatGPT, or heaven forbid, Copilot. It’s been a quarter or two, some cool dashboards have been built and shared, and some emails are being written faster. Perhaps you’ve even started to summarize most meetings using these tools.

At a recent Executive Intensive I led, three of the CEOs attending signed up for their first paid AI subscription just to start the program. These are accomplished leaders running real businesses, and I don't share that to embarrass anyone. It's just where many leaders are starting today.

What does "nothing changing" actually look like?

From the outside, it looks like progress. You get the subscriptions, you tell people to use AI, and somebody publishes a deck of wins. Everybody nods and feels pretty good about it. But if you actually look at the daily workflows inside your business, none of it quite fits.

As your board, the media, and that 22-year-old fluffy haired kid on Instagram have been shouting “use AI” for a while now, most leaders feel a deep pressure to get started. And treating this moment like any other tech inflection point makes sense, as these tools are inherently pieces of software. So we delegate the decision of which one and what access to our IT groups, we wrestle through boring security and governance conversations, and eventually the subscriptions are paid for.

And yet, if a task took five steps before the rollout, it still takes five steps a month later, and five steps the month after that. AI may have automated one piece of the puzzle, but it hasn’t augmented the workflow to be inherently different.

Gallup sees the same gap. In its February 2026 survey, 65% of U.S. employees at organizations that have implemented AI said it had improved their productivity. By May 2026, only 14% strongly agreed it had transformed how work gets done. People feel a little faster. The work itself hasn't changed.

Why doesn't giving everyone a license work?

Say you've never ridden a bike in your life. I drive you to Bentonville, Arkansas (mountain-biking capital of America), rent you a $20,000 mountain bike, and take you to the top of a double black diamond trail. You’ve got the best bike money can buy. “I'll meet you at the bottom” are my parting words.

Would it be reasonable to expect you to fearlessly charge down that single-track? Of course not. You have never ridden a bike before, and learning how to ride a bike is a skill developed over time, with repetition, at varying degrees of difficulty. So why has this been the approach of basically every company trying to adopt AI so far? Why have we handed our people the best tools available, put them at the top of the hardest trail, and then waited to see what happens?

The mistake is viewing this moment like any other SaaS subscription. A subscription is something you rent. AI is not another subscription. It is a fundamentally new capability that has never existed before in our day-to-day workforce, and it requires learned skills to reach its potential. A skill that should be a capability built into you and your people through doing your work every day. It should survive every new model release. And it has to be built the way any other skill gets built.

So what is the skill, exactly?

The skill is directing the work instead of doing it: you clearly describe the problem, you show what “great” looks like for the output, and you judge what comes back until it's right. These tools, well-directed with the right information, context and memory (all better words than data), now do the heavy lifting of creation.

Once you reach a certain level in your career, you have likely become unconsciously competent at a set of actions or skills. How you make a decision. What line in the spreadsheet to check first. Why you can ignore two of the priorities in that email but must address the third one immediately. And if I press you on how you know how to do that, almost all leaders grow silent. “I just do” is the most common answer. It’s similar to asking you how to ride a bike. You can find language to describe it, but it will be imprecise.

Building the capability to fundamentally redesign your workflows is simple, but that doesn’t mean it’s easy. This skill requires you to be able to clearly describe the problem. You must be able to show, not tell, what “great” looks like for the output. You must be willing to grapple with the fact that your role is no longer to “do the work.” But you ultimately judge and are responsible for the output. It's never a blind accept. You redirect it until it's right. We call this skill an “Orchestrator” in a company, but because we are so new to this moment, we don’t even have shared language with which to have this conversation in our companies.

Once you’ve developed this skill, the math on your output fundamentally changes. In the time it takes to sit through a 40-minute meeting, a well-designed system alongside a strong Orchestrator could have the slide deck written and designed, a model built, and a landing page already online. Whether those things are any good depends on how well you’ve organized your information and context. That's the skill.

Towards the end of an Executive Intensive, I watch people begin to grapple with the future of their business and their ambition for it, because they start to understand what's possible.

Why don't training sessions and lunch-and-learns fix it?

I get asked to do lunch-and-learns occasionally. Now I have nothing against lunch or learning, but to be really clear, a lunch-and-learn with your marketing team is not going to change the behavior of your company.

The same goes for telling people to use AI. When we sit down with an executive team, one of the first things we say is to stop telling your people to use AI. That's like telling them to go run a nuclear reactor by themselves.

The gap that exists today between your business and your workflows being transformed is not the tool’s abilities. The gap is the humans, and most of them have been given very little to go on in order to navigate this incredibly disruptive moment. In fact, three in four U.S. employees say their organization hasn't communicated a clear AI plan.

What's the difference between automating work and redesigning it?

Your people don't need to be AI experts. They're already experts in their day-to-day work, so that's where we start. We ask a simple question: how does this work actually get done today?

"Well, I email Jim, and Jim emails Susan, and Susan pulls a report, and it takes three days. Then Susan makes it a PDF and I convert it." That's a friction point, and it's exactly what we're looking for.

The instinct is to automate that chain as it stands, the same emails and handoffs, just faster. That's not the goal. The goal is to understand what you're trying to accomplish, like bringing five pieces of data together every Friday in order for the CFO to make a decision on resource allocation, and then redesign the workflow from the ground up in order to reach that objective.

In the real world, it looks like this: a partner at a small private credit fund came through one of our Executive Intensives back in May. Within a few weeks, he'd rebuilt how his firm screens new deals. The first pass now runs around the clock, and within 10 to 15 minutes he gets a yes or no on whether a deal fits the fund's credit box. In his words: "Probably a lot of other companies would have already hired an analyst. We don't need it yet." And because he understands how it works and what outcome he’s looking for, he can set up future tools for his firm instead of outsourcing this work to a third-party consultant. The capability now lives in his firm and his people, not in dependency on someone outside.

Why do so many smart leaders get this wrong?

Most weeks I talk with 3-5 CEOs who are trying to figure this out. Nearly all of them are doing it with no lived experience. They're operating off a brief, a Wall Street Journal article, pressure from their board, or general curiosity. I don't say that as a criticism. It's where most of us started.

There's a meme that gets passed around among CEOs. Who are we? CEOs. What do we want? AI. AI to do what? We don't know. When do we want it? Right now.

It's funny because it's true. We know it matters, and we know there's a there there. We're just not exactly sure what it is yet. So we go find a deck of wins to show we're doing it, and privately, a lot of us are terrified.

What should a leader do first?

My strong encouragement is simply this: build your own capability. Then make these moves. None of them need a budget. All of them need you.

1. Ask one anonymous question of your employees. "What AI tools are you using for work, approved or not? No judgment." You'll learn where your people already are. The tools they're using on their own can be treated as a violation, or as a roadmap.
2. Refuse the delegation. Don't hand this to IT as a tool project. IT's job is to defend and protect, and it was never set up to build this kind of capability. Ask what data it can handle and what controls you need, not whether a tool is secure.
3. Find the friction. Have your leadership team list the places where work gets stuck. It takes about 15 minutes, not a six-month committee. Pick one, like the monthly reconciliation built from seven Excel files, and give yourself 1 hour to see if you can redesign one aspect of it using your preferred AI tool.

The bottom line

If your rollout hasn't changed how work gets done, it doesn't mean AI isn't working for you. It means a skill got treated like a subscription, and that's fixable. The goal isn't more tools. It's people who own the capability themselves, so you're building capability, not dependency.