Most AI programmes sit inside IT. The returns come from redesigning work, and that is a decision IT cannot make. Here is what leaders must own.
The AI update is item seven on the agenda. The IT director takes four minutes, shows a slide with three pilots on it, and reports that two are progressing well. Everyone nods, and the meeting moves on to the quarterly numbers.
The AI update is item seven on the agenda. The IT director takes four minutes, shows a slide with three pilots on it, and reports that two are progressing well. Everyone nods, and the meeting moves on to the quarterly numbers.
Nothing in that exchange is wrong. But no decision was made in the room, and those three pilots will still be pilots at the next meeting. The AI programme is being reported on rather than led. Reporting is what an organisation does with the things it has already handed to someone else.
Nothing in that exchange is wrong. But no decision was made in the room, and those three pilots will still be pilots at the next meeting. The AI programme is being reported on rather than led. Reporting is what an organisation does with the things it has already handed to someone else.
The adoption argument is over. What has not arrived at the same pace is the return.
McKinsey's State of AI, published 25 August 2026 from 1,719 business respondents across 97 nations, finds that 37% attribute at least some EBIT impact to their organisation's AI use, essentially unchanged from 2025. The group attributing 5% or more of EBIT to AI, and reporting significant value from it, accounts for 6% of respondents. That figure has not moved in a year either.
The adoption argument is over. What has not arrived at the same pace is the return.
McKinsey's State of AI, published 25 August 2026 from 1,719 business respondents across 97 nations, finds that 37% attribute at least some EBIT impact to their organisation's AI use, essentially unchanged from 2025. The group attributing 5% or more of EBIT to AI, and reporting significant value from it, accounts for 6% of respondents. That figure has not moved in a year either.
Both are self-reported attributions by business respondents rather than audited financials, so treat them as direction rather than measurement.
Two years of investment, and the share of organisations getting real money out of AI has stayed where it was.
Both are self-reported attributions by business respondents rather than audited financials, so treat them as direction rather than measurement.
Two years of investment, and the share of organisations getting real money out of AI has stayed where it was.
The comfortable assumption is that the 6% bought better technology, hired better people, or started earlier. The same survey points somewhere considerably less exciting. Nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, against just one-quarter of everyone else.
They are also twice as likely to say their senior leaders demonstrate commitment to AI initiatives, and twice as likely to have a defined process for measuring whether an initiative worked.
The comfortable assumption is that the 6% bought better technology, hired better people, or started earlier. The same survey points somewhere considerably less exciting. Nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, against just one-quarter of everyone else.
They are also twice as likely to say their senior leaders demonstrate commitment to AI initiatives, and twice as likely to have a defined process for measuring whether an initiative worked.
Read those three findings next to each other and a pattern appears that has nothing to do with software. The organisations getting a return are the ones where the work itself was changed, by people senior enough to change it, against a number somebody agreed to in advance.
Read those three findings next to each other and a pattern appears that has nothing to do with software. The organisations getting a return are the ones where the work itself was changed, by people senior enough to change it, against a number somebody agreed to in advance.
Everyone else bought the same tools and left the work exactly as it was.
Everyone else bought the same tools and left the work exactly as it was.
A workflow is not a system. It is who does what, in what order, with whose approval, to what standard, and by when.
So redesigning one means somebody stops doing something they have always done. An approval step moves to a different desk or disappears. A role changes shape, a handover gets deleted, and a team that was measured on volume starts being measured on something else. Not one of those changes sits inside a technology function's authority, and not because IT lacks the skill. None of them are IT's to authorise.
A workflow is not a system. It is who does what, in what order, with whose approval, to what standard, and by when.
So redesigning one means somebody stops doing something they have always done. An approval step moves to a different desk or disappears. A role changes shape, a handover gets deleted, and a team that was measured on volume starts being measured on something else. Not one of those changes sits inside a technology function's authority, and not because IT lacks the skill. None of them are IT's to authorise.
Which produces a structural problem that is easy to miss and expensive to leave alone. The function most organisations have made responsible for AI is the one function that cannot approve the change that produces the return.
This is not an argument against involving IT. It is an argument about where the ceiling comes from. An AI programme run as a technology initiative is capped at whatever value can be extracted without changing how anyone works, which is roughly the difference between a faster first draft and a better business. We wrote last week about what has to exist on the Monday after a workshop ends. This is the decision that sits above it.
Which produces a structural problem that is easy to miss and expensive to leave alone. The function most organisations have made responsible for AI is the one function that cannot approve the change that produces the return.
This is not an argument against involving IT. It is an argument about where the ceiling comes from. An AI programme run as a technology initiative is capped at whatever value can be extracted without changing how anyone works, which is roughly the difference between a faster first draft and a better business. We wrote last week about what has to exist on the Monday after a workshop ends. This is the decision that sits above it.
In our own Align sessions the disagreement is almost never about which tool to use. It is about which process goes first, and that argument gets settled by whoever has the authority to stop something.
In our own Align sessions the disagreement is almost never about which tool to use. It is about which process goes first, and that argument gets settled by whoever has the authority to stop something.
This is the work the Align stage of our AI Success Workshop is built around. Leadership teams bring their own candidate projects and score them on the AI Impact Matrix, impact against feasibility with the data and the people they actually have, until the room can repeat one prioritised list back without arguing.
This is the work the Align stage of our AI Success Workshop is built around. Leadership teams bring their own candidate projects and score them on the AI Impact Matrix, impact against feasibility with the data and the people they actually have, until the room can repeat one prioritised list back without arguing.
Twenty minutes, no budget, no vendor. Take the AI initiative your organisation is furthest along with, and answer four questions out loud.
Twenty minutes, no budget, no vendor. Take the AI initiative your organisation is furthest along with, and answer four questions out loud.
Four answers, one meeting. If three of the four have no name attached, the initiative is not under-resourced. It is unowned, and no amount of additional technology will fix that.
Four answers, one meeting. If three of the four have no name attached, the initiative is not under-resourced. It is unowned, and no amount of additional technology will fix that.
The gap between the 6% and everyone else is not a technology gap. It is the distance between an organisation where AI is reported on and one where somebody senior has decided what changes, who agrees, and what number it moves.
The gap between the 6% and everyone else is not a technology gap. It is the distance between an organisation where AI is reported on and one where somebody senior has decided what changes, who agrees, and what number it moves.
That decision cannot be delegated downward, and it does not get easier by waiting. Organisations transform when leaders transform first.
The Align stage of the AI Success Workshop puts your leadership team in a room with your own candidate projects and works them down to a prioritised AI Impact Matrix: the ranked shortlist, with the first process circled and the number it should move written next to it. Workshops run across Asia, with dates by region.
Copyright © 2026 Vision Academy. All rights reserved.
