Nearly nine in ten organisations now use AI. Stanford's 2026 AI Index puts the figure at 88% for 2025.

Yet in McKinsey's 2026 survey of 1,719 executives, only 37% could point to any impact on earnings, the same share as a year ago. Just 6% reported significant value.

That isn't an adoption gap. It's a design gap, and it is widening.

Adoption is no longer an advantage. It's the baseline. The companies pulling ahead aren't using more AI. They have rebuilt how work gets done around it. And in 2026, for the first time, you can see the difference from the outside.

Three bars narrowing from 88% of organisations using AI, to 37% seeing any impact on earnings, to 6% seeing significant value.
Almost everyone uses AI. Very few get real value from it. Sources: Stanford HAI, AI Index 2026 (2025 data); McKinsey, The state of AI in 2026. The figures come from separate surveys.

Three levels of AI use

We talk about "using AI" as if it were one thing. It is three.

  1. Assist. Individuals use AI tools to write, code and summarise faster. The gains are real, but they are personal.
  2. Automate. AI performs one step inside an existing process. Same process, lower cost.
  3. Redesign. The workflow is rebuilt end to end. Agents do the work. People set direction and handle exceptions. Speed, capacity and economics all change.
A three-step staircase: 1. Assist (changes a person), 2. Automate (changes a step), 3. Redesign (changes the whole workflow).
Automating makes you cheaper. Redesigning makes you faster than your market.

Most companies are stuck between one and two, and the numbers show why it doesn't reach the P&L. Individual gains, however widespread, don't add up to company results on their own.

The companies that do see results work differently. Nearly three-quarters of McKinsey's high performers have fundamentally redesigned their workflows. Among everyone else, it is one in four.

Two bars: about 75% of AI high performers have redesigned their workflows, against about 25% of everyone else.
The difference is redesign, not usage. Source: McKinsey, The state of AI in 2026.

Take customer onboarding at a software company (an illustrative case). Automating it means AI drafts the kickoff email. Redesigning it means an agent provisions the account, configures the integrations and flags only what it can't resolve, so onboarding drops from weeks to days. The first makes your team a little faster. The second changes how many customers you can take on.

Automating tasks makes you cheaper. Redesigning workflows makes you faster than your market.

The gap compounds

The tempting response is to wait until the dust settles. Waiting is not neutral.

BCG's 2025 study of 1,250 executives found that the 5% of companies it calls "future-built" plan to spend more than twice as much on AI as laggards. Compared with those laggards, they report 1.7 times the revenue growth and 3.6 times the three-year shareholder return. The likely mechanism is reinvestment: leaders put their gains back in, and the distance grows.

Paired bars comparing future-built leaders with laggards: 1.7 times the revenue growth and 3.6 times the three-year shareholder return.
Leaders pull away on growth and shareholder returns. Source: BCG, The Widening AI Value Gap (2025). Self-reported survey data.

The sceptics have a point, and it deserves a straight answer. McKinsey's headline number hasn't moved in a year. Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027, citing costs, unclear value and weak risk controls. It also warns of "agent washing", where existing products are relabelled as agentic.

Both things are true. A lot of AI spending is going nowhere, and the few who redesign properly are pulling away. The lesson isn't "do more AI". It is "redesign the right work, with discipline".

What the leaders do differently

Four habits show up again and again.

  1. They start from throughput, not tasks. They pick two or three workflows that set the pace of the business, such as onboarding, support resolution, the release cycle or quote-to-cash. Then they ask what each would look like rebuilt, not sped up.
  2. They aim for growth as well as cost. McKinsey found high performers are far more likely than others to intend to use AI to fundamentally transform their business. A cost target caps ambition at what you already spend. A growth target asks what you could do that you can't do today.
  3. They measure outcomes, not usage. Cost per resolved ticket, days to go-live and releases per quarter tell you whether work changed. Licences and tokens only tell you that people logged in.
  4. They look outwards on a schedule. Each quarter, they check the public footprints of the nearest AI-forward competitor (its changelog, API docs and pricing page), the way they already track pricing and win rates.

The question to take into Monday

Adoption is now the baseline. Redesign is the difference. And that difference is starting to show up in places other people can see.

So here is the question I'd put to any leadership team this quarter: can you name one workflow you have redesigned around AI, not just sped up? I'd like to hear yours in the comments.

Sources