85% use AI. 29% changed how they work.
This page is about why AI access does not become AI use, and why use does not become a changed way of working. It covers the staircase most rollouts stall on, four situations where that stall shows up, and the four-step process we use to get from access to an actual change in how a team decides and works.
30 minutes. No pitch. No follow-up pressure.
Why access never becomes a changed way of working
An AI license changes what appears on someone's screen. On its own, it does not change how a decision gets made in the room. Most rollouts stop at the first step of a four-step staircase: access, then occasional use, then a genuine change in daily work, then a measurable change in what the team actually produces. Most organisations get people to step one. Almost none reach step three.
The instinct after low adoption is more training, more prompts, more use-case examples. Those raise step two. They do not touch step three, because step three is not a skills gap — it is a decision gap. Nobody in the room owns the decision to actually change how the work gets done.
What we see most often
The workaround economy
Six months after the AI tool launched, the old process is still running in parallel.
Not resistance — necessity. The tool does not handle a specific exception that comes up forty times a day. Nobody told the team that built it. The workaround became the real system.
We have unblocked this.
The excluded expert
The people who know the process best were never asked before the tool was chosen.
Frontline staff have four objections that would have changed how it got rolled out. They raised them once, in a survey nobody read. Now they use the tool for the easy parts and route around it for everything that matters.
We have unblocked this.
The middle manager veto
Leadership approved the rollout. Team leads did not.
The tool is technically available to everyone. The team lead has a different way of working that keeps their existing reporting intact. Their team follows the lead, not the official process.
We have unblocked this.
There is no decider
Everyone agrees the AI tool should change how the team works. Nobody has actually decided what changes.
The rollout has an owner for the software. It has no owner for the decision — which parts of the old process get retired, which judgment calls the tool now makes, who signs off when it is wrong. Without that decision, people keep doing both: the new tool and the old safety net.
We have unblocked this.
How we close the gap
Adoption diagnosis
We talk to people using the AI tool and people who quietly stopped, separately. We find the specific moments where they revert to the old way, and why.
Blocker session
We bring together the people who own the rollout and the people working around it. Structured methods surface the real objections, not the polite version.
The decision session
For each blocker, and for the decision nobody made yet — a call, an owner, a timeline. Not a promise to investigate. A commitment made in the room by the person responsible for keeping it.
Adoption plan with teeth
A 30-day plan built by the people who will execute it. Specific actions, specific owners, specific check-in points. Documented before anyone leaves the room.
When to call us
- People have access to the AI tool but almost nobody has changed how they actually decide or work
- The same objections keep surfacing informally and nobody owns fixing them
- Nobody can say who decided which parts of the old process the tool replaces
- Nobody has asked whether people trust the tool's output enough to act on it without double-checking
- Adoption numbers only track logins and usage, never what got slower or worse because of the rollout
A promise about who sits at the table
Our sessions are designed so that the night shift, the non-native-language colleague, and the person who does not enjoy writing contribute as much as anyone else. Not as an afterthought — as a design principle. The people who are least often in the meeting room have the strongest voice in the room when we are done.
- The night shift — workers on rotating shifts get formats that fit how they think, not how the meeting room is set up.
- The non-native-language colleague — contributes in whichever language is sharpest for them, without the translation lag that dulls the thought.
- The person who does not enjoy writing — visual formats, drawing and building make thinking visible without anyone having to type up a summary.
See also
Sources
Each source below underpins a mechanism discussed on this page.
- Access does not become use, and use does not become a changed way of working -- the same pattern that stalls scaling after a pilot. Haefner, N., Wincent, J., Parida, V. & Gassmann, O. (2023). Implementing and Scaling Artificial Intelligence. Technological Forecasting and Social Change 191, 122878. DOI 10.1016/j.techfore.2023.122878
- Fast adoption, slow uptake -- the barriers are skills, data, cost and organisational readiness, not access to a model. OECD/BCG/INSEAD (2025). The Adoption of Artificial Intelligence in Firms: New Evidence for Policymaking. OECD Publishing, Paris. DOI 10.1787/f9ef33c3-en
Where the AI expertise lives
This page is about the conversation and the decision an AI rollout exposes — not about picking the tool or designing the workflow around it. For AI governance and board-level strategy: aiexecutive.be. For hands-on use-case design and workflow redesign: agentic-engineering.be.
Book a discovery call
Tell us which tool, what the adoption numbers look like, and what you have already tried. We will ask questions, listen, and tell you honestly whether the blocker is a conversation we can help you have. If it is not, we will say so.
30 minutes. No pitch. No follow-up pressure.