When AI Makes Activity Cheap, the Outcome Is the Only Moat

$2,000 a month. That is what some of Uber's 5,000 engineers were spending on AI coding tools.

The company used its entire 2026 AI budget in four months.

The tools were not failing. They were working, so people used them more and more.

It is the kind of story that gets shared as a cautionary tale about AI cost. That misses the deeper lesson.

The half they could see

Uber are not careless. They are one of the most respected, both commercially and technically adept businesses in the world.

But they measured the half they could see.

The cost was visible to the penny. The value was not.

The line from a line of code, to a shipped feature, to a faster pickup, a lower cost per route, a rider who comes back, was never built.

So they rewarded usage instead. On a leaderboard. Engineers were ranked by how much AI they used.

Usage is the input, not the result. The spend followed the thing they measured.

Then Uber's COO was asked the question that closes the loop. Was all that AI spending producing more for customers? He was honest. He could not yet draw the line. The connection, he said, is not there yet.

That is the Attribution Gap, said out loud by the COO of one of the most AI-forward companies on earth.

What it means for the rest of us

If Uber, with their data teams and their resources, can fall into this, what chance does a £20m service business have?

Microsoft has wound back its own internal AI coding pilot, six months in. Forbes has flagged it as the start of a pattern, not a one-off.

It is not a competence problem. It is a structural one.

AI has changed the shape of cost. Old software licences were a fixed price. You bought the seats. The number did not move with use. There was nothing to control.

Token pricing broke that. Now the cost tracks the usage. The moment the tool gets popular, the cost climbs with it. There is no ceiling.

Spend and output have become the same curve. Almost no one can yet tell which pound moved the dial.

This is the new gap, and it is coming for every business that uses AI.

And not just for AI spend. If your last renewal conversation was about hourly rates, you are already aligned to activity. The same gap, on a different invoice.

What it looks like when you flip it

One service business we work with had a contract running on the standard model. Cost-plus pricing. 30 per cent gross margin. Gross profit of £75,600 a year.

Nothing wrong with the work. The customer was happy. The team was meeting every service level.

They did not add new logos. They did not hire. They did not change the technology.

What they did was build the line. From the operational work they were already doing, to the outcomes the customer actually cared about, using the data already in the business. They showed the customer what their service had caused. The customer agreed, and the contract repriced on the outcome.

Gross profit on that same contract moved from £75,600 to £184,350. A 144 per cent increase. Margin moved from 30 per cent to 46.

The work did not change. The proof changed.

That is the difference between aligning to activity and aligning to the outcome, on a single contract, in real money.

Outcome-led, the approach

This is what being outcome-led fixes, it is about aligning the whole business, it is a mind-set, not a slogan.

You start from the outcome the customer actually wants. In Uber's case, a faster pickup, a lower cost per route, a rider who comes back. For a services business, the same logic, different units: time-to-value for the client, net retention on the book, gross margin defended at renewal.

You baseline it. From real data, in the business as it is today. The retention as it stands. The time-to-value as it actually is. The margin on the existing book this quarter.

You decide which things, people, software, AI spend, are meant to move it.

You build the line that shows what each one produced. In practice that means going into the data the business already holds, the tickets, the logs, the financials, and showing which operational actions caused which customer outcomes, by how much, over what period.

Then you run and adjust against the outcome, not the activity.

That is Outcome Engineering. Four steps. No theory. Done with the data the business already has. 🎯

How it aligns the whole business

The point of the approach is alignment.

The whole business points at the same result. Every team, every tool, every pound of spend is judged by the outcome it produced, not by how busy it was.

Incentives flip. The leaderboard, if you must have one, ranks people on outcome contribution, not usage.

Decisions flip. They start from the result that has to land, then work back to the people, the software, the AI spend, the process change. Not the other way round.

Budgets flip. Spend can no longer run away, because every pound is sitting next to the outcome it bought.

When that is in place, AI becomes a multiplier, not a leak. Usage rises because it is producing more, and you can see it.

Why this is a moat, not a slogan

The AI itself is on a shelf for every competitor by next quarter. That is not the moat.

The line from work to outcome is the moat, because it cannot be bought off the shelf. It is built from the inside, on the business's own historical data, contract by contract. It takes months of patient work, and a different way of thinking, what changed for the customer, not what we did.

Most competitors will not do it. The result is invisible until it lands, and the work does not look like progress while it is happening. So they stay on activity, and compete on the price of it.

The ones who do build the line compound. Renewals lift. Margin lifts. The exit multiple holds.

Why this matters now

A year ago, outcome-led was a useful edge. The businesses that had it grew slightly faster and renewed slightly higher.

Now it is the whole game.

AI is about to flood every business with cheap activity. The activity is no longer the moat. Anyone will have it. The only defensible position left is being able to show what the activity produced. For your customer. For your margin. For your enterprise value.

A business that can show this can price on the result, defend the multiple, and grow what it already has. A business that cannot, will compete on the price of the activity. And the price of activity is heading toward zero.

Uber aligned to usage. The lesson is to align the business to the outcome.

So a fair question to end on. If you had to pick one customer outcome and align the whole business to it, what would it be?