Your team got faster this year. The business didn’t.

That's the story almost everywhere, the work just got quicker. A few teams used the same emerging tools to do things they couldn't do before, and the difference between faster and new is the only one that matters.

Walk into almost any company right now and AI is everywhere and nowhere at once. Your people use it all day, writing the first draft and tightening the deck and turning around in minutes what used to take a week. Everything ships faster. And yet how the business actually works, where growth comes from and how the real decisions get made, looks about the same as it did for the last decade. That's the gap you can feel but haven't quite been able to name.

Because there are two very different things AI can do for a business, and they're easy to confuse. AI can help you do the work you already do quicker and more efficiently. Or it can change what the business is capable of. The first gives you the same playbook at lower cost. The second gives you capabilities you didn't have before, work and outcomes that weren't available at all. Those aren't two speeds of one thing. They're different intentions, and they build very different companies.

Notion's AI maturity model

Notion's 2026 research surveyed 6,118 AI decision-makers and users across ten markets. Its public summary places 88% of organizations in the early stages of adoption. It describes a shift from using AI for speed toward building new capabilities as adoption deepens. The framework below is a useful way to examine that distinction; it is not a causal test of why any one company advances.

The four stages of AI maturity. Find where your function or business unit sits. Then decide to cross the line.

1. Thought Partner. People draft and brainstorm on their own.

2. Assistant. AI draws on your data and helps with real tasks.

The line that matters. Stages 1 and 2, AI does your work. Stages 3 and 4, it does work you couldn't.

3. Teammate. Agents run recurring workflows; people review them.

4. System. AI supports critical work across connected processes.

Notion Global AI Transformation research, 2026. Survey of people describing their organizations, not an audit of 6,118 companies.

Almost everyone sits in the bottom half of that curve, in stages one and two, below the line. More than half are at the first stage, using AI the way your team probably did this morning, as a smart helper on the task in front of them. Only a few have crossed into stages three and four, where AI runs the work itself. The study is pretty direct about why that feels surprising: the loudest stories come from the edges. A small handful of companies are doing the disruptive work, loudly enough to convince everyone else they're falling behind. Most of the market isn't behind at all. It's just below the line.

The public summary identifies a shift in intended value: from speed toward new capabilities. My read is that leadership has to make that intention concrete. What should the team become able to do, and which part of the work needs to change?

Below the line, AI makes your team faster. Above it, it does work your team never could.

Who's allowed to cross the line

Using AI to make an existing task faster can start with one person. Changing what a function produces usually needs authority over its goals, resources and workflow. The survey can’t prove that permission is the reason organizations advance. It is where I would start the conversation with a leadership team.

But permission only explains who can cross the line. It doesn't explain why so few of the people who can, actually do. The honest reason is harder to say out loud. Making the existing work more efficient is the safe move: it asks for no vision and risks nothing, and you can do it for years and look completely responsible the whole time. Changing what the work is means committing to a view of where things are headed and being the one who's wrong if it doesn't land. That exposure is the real barrier, and it isn't technological. The distance between the few who cross and everyone who doesn't was never mainly about tools, or even strategy. It's a courage gap. AI didn't create it. It just stripped away the last excuse for not facing it.

Look at what that means for your business or function. If your team is stuck below the line, it isn't because they aren't using AI. They use it constantly, to move faster. It's that moving faster is the most they can do without you. The ceiling on what AI actually changes in how you grow is you, the most senior person willing to rethink how the work gets done.

It was never a technology gap. It's a courage gap.

The scaffolding beneath the crossing

The barrier isn't only courage. There's a practical one beneath it that trips up even teams with the will to cross. The everyday tools that got you this far, a ChatGPT, Copilot, or Claude, are probabilistic: brilliant guessers, which is exactly what you want for a draft you're going to read and fix. That's stages one and two. But running real work end to end needs something a raw model can't give you on its own, reliability you can trust without standing over it. That comes from the deterministic scaffolding built around the model, the guardrails and expert prompting that turn a good guess into a dependable system. Most teams never see that gap, which is why a chatbot subscription gets mistaken for transformation, and the confusion runs all the way up, from the person typing the prompt to the C-suite signing the budget. What actually crosses the line looks less like a chatbot and more like a harness around the model. Something like FifthRow: expertly prompted, model-agnostic, built to run the work rather than just help with it.

We've seen this exact lag before. When factories first got electricity, most owners did the obvious thing and swapped the steam engine for an electric motor. They got a modest bump and wondered what the hype was about. The ones that actually won spent years on something harder. They redesigned the whole floor around what electricity made newly possible, and changed what the factory could make. The motor was available to everyone. The redesign was a leadership choice, and it took a generation. AI is the same pattern, running faster. The tools are already everywhere. The decision to rebuild how the work gets done is, as it always has been, the rare part.

So you're not behind on AI. The question was never whether your team is using it. It's whether you'll move your business or function across the line on purpose, and that's work you can start this quarter. It begins with being honest about why the team is stuck below it, because the reasons are rarely about the tools.

Crossing starts the moment you point AI at something it has never touched in your world. Have your team use it to learn a new domain in days instead of months, so missing expertise stops being the thing that holds you back. Turn it loose on the bigger problems worth solving, not just the work already in the queue. And aim it at the real areas of disruption you see coming, the ones that actually scare you, instead of the safe edges everyone else is automating. That's what above the line looks like in practice. Getting your whole team there takes a few deliberate moves.

How you cross the line

Four moves, and every one of them is yours to make.

1. Show them what's possible. Most teams can't build a stage-three capability because they've never seen one. Their picture of AI stops at a faster way to write or research, so that's the ceiling they aim for. Don't tell them to think bigger, show them. Pick one company already operating above the line and pull it apart together: what did they actually build, and what would our version of it be? Bring in a peer who's crossed and have them walk your people through how they did it. And get one small proof on the table fast, have someone wire up a single agent to run one real workflow, then demo it, because nothing widens the picture like watching it work. Imagination is the first input, and it's the cheapest one to raise.

2. Make the people plan explicit. Explain the intended effect on roles and responsibilities honestly. If the goal is to direct recovered time toward work the team has not had capacity to pursue, name that work and fund it. People need to understand the plan before they can help redesign the process.

3. Get fluent fast at one and two. Mastery of the bottom funds the top. The quicker your team gets genuinely fluent with AI on today's work, the more time and confidence it has to build tomorrow's. So set a real bar, a standard everyone is expected to hit rather than a suggestion to dabble, and give people an easy way to share what's working so the best moves spread fast. Push proficiency at the first two stages hard, not as the destination, but to clear the hours and the headspace that stage three and four demand.

4. Make the portfolio visible. You can't deliberately move what you can't see. Start simple: list every place AI is being used across the team and mark which stage each effort sits at. The moment it's on one page, the imbalance is obvious, and you can shift the mix toward three and four on purpose instead of by accident. That visibility is what turns four good intentions into a plan.