Buying AI Tools Isn't an AI Strategy. Building Capability Is.

Most businesses have an AI strategy that fits on a single slide. A list of tools they've paid for. ChatGPT Teams. Copilot. A Notion AI subscription somewhere. Maybe a Claude Enterprise seat for the executive team.

Next to each one, a number. The monthly cost. The seat count. The renewal date.

That isn't a strategy. That's procurement.

I see this everywhere right now. Boards approve AI spend, dashboards get built, vendors get paid, and twelve months later somebody asks the obvious question. What are we actually getting back?

The answer is usually uncomfortable.

The Gap Between Spend and Value Is Widening

Reports from the last six months keep telling the same story. Companies are spending more on AI than they were a year ago. The percentage reporting measurable value from that spend hasn't moved much. In some surveys it's actually gone backwards.

That isn't because the technology stopped working. The technology kept getting better. The frontier models in May 2026 are dramatically more capable than what was on offer eighteen months ago.

So what's missing?

I think the answer is the unglamorous one. Capability. The actual ability of the humans inside your business to think with these tools, direct them well, review their output, and weave them into how the work gets done.

Tools don't create value. People using tools well create value. The first part is buyable. The second part is not.

What “Capability” Actually Means

When I say capability, I'm not talking about prompt cheat sheets passed around in a Slack channel. That's the version most companies stop at. A two-hour training session, a PDF of prompt templates, a self-congratulatory email about “AI rollout complete.”

Real capability is layered. Three levels worth naming.

Literacy. The basic understanding of what these systems are, what they're good at, where they fail, and how to talk to them. Every person in the business who touches knowledge work needs this. It isn't optional anymore.

Fluency. The working skill of using AI inside real day-to-day tasks. Drafting, summarising, analysing, structuring thinking, getting unstuck. This is the level where productivity gains start showing up in the numbers. Most people get here within a few weeks if they actually practise on real work.

Leadership. The ability to redesign workflows around AI, decide where it belongs, where it doesn't, and how the team should be set up to use it. This sits above the tool. It's strategic, and most businesses don't have a single person operating at this level yet.

If your AI rollout stopped at literacy, or hasn't even reached there in a serious way, you don't have an AI strategy. You have a software subscription.

Why This Gap Is So Easy to Miss

I see this pattern over and over for two reasons.

Buying is easy. Capability building is hard. A licence is a credit card. A capable team is a year of patient work. Boards reward the first, because it shows up cleanly in the next quarterly slide. Nobody gets promoted for slow, distributed capability gains across forty people.

The symptoms of low capability are also quiet. They look like normal mediocrity. A report that took two days when it could have taken two hours. A draft email written from scratch when a strong starting point was three prompts away. A junior analyst spending an afternoon on a manual cleanup task that AI would have done in fifteen minutes.

None of that shows up in your AI dashboard. Your AI dashboard shows that everyone has access. That's it.

Meanwhile, your business is running at a fraction of the pace it could be running at. You just don't see it, because there's nothing visible to compare it to.

What Good Looks Like

The businesses I work with who have actually built capability tend to share a handful of patterns.

They invest in real training that goes past the first hour. Not a single webinar. A genuine programme that runs for weeks, with practice, feedback, and people building their own use cases against their own work. The kind of training where a sales rep ends up with three or four AI workflows they own and use every day, not a folder of templates someone else made.

They have someone senior accountable for capability, not just for tool selection. This is often a Head of AI or an equivalent role inside the leadership team. Their job isn't vendor management. Their job is making the humans in the business better at this.

They tolerate visible learning. People are encouraged to share what's working. Failures get written up rather than buried. There's a regular cadence, usually monthly, where teams demo what they've built, what's saving them time, and where they're stuck. This is where culture starts to compound.

They make it part of how performance gets judged. The intent isn't punitive. The expectation is that being capable with AI is part of being good at your job, the same way being capable with email or spreadsheets has been for the last two decades.

The Amplifying Intelligence Frame

I keep coming back to a simple framing for this. AI doesn't replace human intelligence. It amplifies it. The more intelligence is already there, the more the amplification compounds.

Two analysts have the same access to the same AI tool. One has spent six months building genuine fluency on real work. The other watched a webinar last year. The first one is doing work the second one can't even see as possible. They aren't using different tools. They're operating at different levels of capability.

That gap is what's quietly deciding which businesses pull ahead in this cycle and which get left behind. It isn't visible from the licence count. It's only visible from the output.

This connects to a question I've written about before. The ROI signal you're looking for shows up here, not in the procurement line. Read more on that in the piece on how to measure AI ROI. The shadow AI conversation also lives in the same place, because shadow AI is, accidentally, capability building happening without anyone leading it.

A Practical 90-Day Capability Plan

If any of this is uncomfortably familiar, here's where I would start in the next quarter.

• Map current capability honestly. Most businesses overestimate where their team is. A short skills assessment, focused on real tasks rather than self-rated confidence, will tell you the truth. Pay attention to the gap between the executives' assumptions and the team's actual day-to-day behaviour.
• Pick one workflow and go deep. Don't try to roll out AI fluency across the whole business at once. Pick one workflow with one team, train them properly, build genuine fluency in that area, and use it as the proof point. Capability spreads better from a strong example than from a thin programme stretched over everyone.
• Run a monthly demo cadence. Once a month, a different team shows what they've built or learned. Twenty minutes, no theatre. This is the single highest-leverage cultural move you can make, and it costs almost nothing.
• Name someone accountable. Even if the role is part-time. Someone has to wake up thinking about whether the business is getting better at this, week over week. If nobody owns it, it doesn't happen.

This is unglamorous, slow work, and it's exactly the work most businesses skip because it doesn't fit neatly inside a quarterly plan. I would massively challenge any leader reading this to spend more in the next twelve months on capability than on additional licences.

You Can't Buy Your Way Out of This

For me, the closing thought is straightforward. There is a version of the next five years where the businesses that win aren't the ones that bought the most AI. They're the ones who built the most capable humans around AI.

That second category is mostly invisible right now. It doesn't show up in vendor announcements or earnings calls. It will show up in three years, when the gap between the capable businesses and the rest becomes impossible to miss.

Tools are commodity. Capability isn't. You can't buy your way out of this one.

If you'd like help thinking through what a serious capability plan looks like in your business, that's work I do through AI consulting and structured business coaching. If you want a quick sense of where you actually are, the quiz will get you to a sensible first move in a few minutes.

Frequently Asked Questions

What's the difference between AI tools and AI capability?

Tools are the platforms you pay for, like ChatGPT, Claude, or Copilot. Capability is the actual ability of the humans inside your business to use those tools well, direct them, review their output, and build them into how the work gets done. Tools are buyable. Capability is built over time, through deliberate training, practice, and leadership.

How long does it take to build real AI capability in a team?

Basic literacy can be established in a few weeks if the training is serious. Working fluency, where someone actually integrates AI into their real day-to-day tasks, usually takes two to three months of guided practice. Genuine team-level capability, with documented workflows and shared standards, is a six to twelve month build. There are no shortcuts that hold up under scrutiny.

Who should be accountable for AI capability in a business?

Someone senior, named, and resourced. The title varies. Head of AI, Director of AI Enablement, Chief AI Officer in larger businesses. What matters is that one person wakes up thinking about whether the business is getting better at this, week over week. If responsibility is spread across IT, HR, and operations, nothing meaningful happens.

Why isn't a one-day AI training session enough?

Because capability is a practice, not an event. A one-day session can plant useful concepts, but skill comes from repeated use on real work, with feedback. The businesses I see making genuine progress run training over weeks or months, with people building their own use cases and sharing what they're learning. A single workshop without follow-up is a budget line item, not a capability programme.

How do I know if my team has real AI fluency?

Look at the work itself rather than the dashboards. Do people use AI for real tasks without being prompted? Can they describe specific workflows they've built? Are they catching weak AI output and improving it rather than just shipping whatever the model gave them? If you ask three people in different roles to show you how they're using AI this week, and you get specific, confident, varied answers, you have real fluency. If you get vague gestures, you don't.