AI adoption is rising. Why is business impact still uncertain?

Every survey tells the same story: AI use is up, and up sharply. Stanford's AI Index charts steep rises in adoption and capability, and the 2024 Work Trend Index from Microsoft and LinkedIn, a vendor survey, found three in four knowledge workers already using AI at work. Then comes the quieter boardroom question that no chart answers: so where are the results?

That question is fair, and the honest answer is that adoption has outrun impact. The MIT Media Lab's GenAI Divide report found high adoption sitting alongside very little measurable business return. Its headline figure has been contested and should be read as a provocation rather than a precise measurement, but the shape is real and most leaders recognise it. The gap between using AI and benefiting from it is not a mystery. It is a ladder, and most organisations have climbed only the first rung.

The adoption-impact gap, rung by rung

The reason "we use AI a lot" and "AI has changed our results" can both be true is that value has to climb four distinct steps, and a different thing blocks it at each one.

Access without competence. Everyone has the tools, but few use them with judgement. Logins are not literacy, and a workforce with accounts it cannot use well has adoption on paper and nothing underneath it.

Competence without workflow integration. Some people get genuinely good with AI, but the workflow around them is unchanged, so the skill stays personal and never compounds. A brilliant individual habit that the process does not support dies when that person is busy, or leaves.

Integration without measurement. The workflow changes, but nobody measures whether it helped. Value that is not measured cannot be seen, defended, or repeated, so it stays a matter of anecdote and faith rather than something the business can build on.

Measurement without strategic reinvestment. You can finally see the saving, and then it leaks away, because the freed capacity is not deliberately redirected at anything that matters. Time saved that quietly refills with more of the same work shows up nowhere in the results.

Each rung is a different bottleneck, and they have to be climbed in order. This is also why the lag is real rather than imagined: Brynjolfsson, Rock and Syverson's productivity J-curve describes exactly this, the payoff from a general-purpose technology arriving only after the unglamorous complementary work is done. Access is instant. The four rungs take time.

Access raced ahead. Results are still waiting for the work in between.

Why this reframes the problem

Most organisations believe they have a vague, single "AI problem". They almost always have a specific rung problem, and naming it is most of the cure, because each rung has a different fix.

A business stuck at the first rung does not need more tools; it needs AI literacy and a sense of where AI is actually worth using. One stuck at the second has skilled people and unchanged processes, and needs the workflow redesign that training alone never delivers. The third needs to start measuring genuine business value rather than activity. The fourth needs the strategy and reinvestment that turns a saving into an outcome. The wrong fix for your rung is wasted effort, which is why so much AI spending feels busy and changes nothing.

What this looks like in practice

Consider a professional services firm with enthusiastic, high adoption and a leadership team quietly puzzled that none of it shows up in the numbers. A short diagnosis finds the firm stranded between the second and third rungs: plenty of skilled individuals, almost no changed workflows, and no measurement at all. The instinct had been to buy another tool. The actual need was to integrate one workflow properly and measure it. The tools were never the problem; the rungs above them were. (An illustrative example, not a specific firm.)

What makes the ladder worth taking seriously is that the rungs compound. An organisation that builds genuine literacy finds use cases easier to spot; one that has integrated a workflow finds measurement natural; one that measures honestly finds reinvestment obvious. The work at each rung makes the next one cheaper, which is why the distance between the organisations getting value and the ones merely using AI tends to widen over time rather than close. It also explains why buying more tools so rarely helps: a new tool adds to the bottom rung, where almost everyone already has plenty, while the real bottleneck sits two rungs higher. The leaders pulling ahead are not the ones who bought the most. They are the ones who treated adoption as a sequence to climb rather than a purchase to make.

The honest limits

Two cautions. First, because the lag is structural, a lack of measurable impact today is not proof of failure. The J-curve warns against killing a sound effort just because the spreadsheet has not caught up yet. But lagging is not the same as imaginary, and the discipline is to keep climbing rather than to use the lag as an excuse for never measuring at all.

Second, the ladder is a way to think, not a rigid sequence. In practice the rungs overlap, and a large organisation can sit on different rungs in different teams. The value is diagnostic: it turns "AI is not working for us" into "we are stuck at rung two in the client team", which is a problem a leader can actually act on.

What to do next

Find your rung, honestly. Ask whether your people are genuinely competent or merely equipped, whether any workflow has actually changed, whether anything is being measured, and whether the freed capacity is going anywhere on purpose. The first of those questions to get an uncomfortable answer is your rung. Fix that one, not the one a vendor would prefer to sell you.

The tool

To place yourself honestly, we have built the AI Adoption Maturity Ladder: a short diagnostic that locates your organisation across the four stages, from initial access to measurable, reinvested impact, with the recommended next action for each rung so you spend effort where it will actually move you up.

Download the AI Adoption Maturity Ladder (PDF)

Working out your rung and the next move is precisely what an AI Reality Check with a team is for, and building a leader's own judgement about all of this is the job of our AI lessons for leaders. It is also the thread running through this whole series, from literacy at the bottom of the ladder to reinvested value at the top.

Sources and further reading

• Stanford HAI, 2025 AI Index Report. Independent. Source for the steep rise in AI adoption and capability.
• Microsoft and LinkedIn, 2024 Work Trend Index. Vendor survey. Source for three in four knowledge workers using AI at work. Read as vendor research, not independent measurement.
• MIT Media Lab (Project NANDA), The GenAI Divide: State of AI in Business 2025. Industry report, not peer-reviewed; the headline 95% figure has been contested. Source for high adoption alongside low measured impact.
• Brynjolfsson, Rock and Syverson, The Productivity J-Curve, NBER Working Paper 25148, 2018. Independent. Source for the structural lag between technology adoption and measured value.