Change management: from delivered to adopted
A tool that has gone live is not a tool that has been adopted. Across roughly 200 projects, the gap between the two has almost always come down to usage, not technology. I treat it as a project deliverable, and because I understand what the tool really does, I know what it changes in each person's work.
Why a delivered project can still fail
A project can meet its deadlines, budget and scope, and still fail: the tool is live, but teams keep working around it, in their spreadsheets, their emails and their habits. That is not a technical problem. It is an adoption problem, and it is prepared long before go-live.
Studies point the same way. According to Prosci, a project whose change is managed really well is seven times more likely to meet its objectives. In France, only 18% of companies with 10 or more employees used AI in 2025, and 61% of people who use it at work have had no training from their employer.
Resistance to change is not a flaw in users. It is a normal reaction when a tool changes the way people work without anyone explaining why, or listening to what the people on the ground lose in the process. AI makes this even more visible: an assistant changes the content of the work itself, not just the tool that carries it.
Sources: Prosci, Best Practices in Change Management, 12th edition (2023): 88% of projects with excellent change management met their objectives vs 13%, self-reported data ; Insee Première no. 2120 (2026): companies with 10 or more employees, 2025 ; Microsoft and LinkedIn, Work Trend Index (2024): 39% trained by their company
My method
I treat change as a workstream of the project, with its own scoping, schedule and indicators, not as a training session tacked on in launch week.
Analyse the impacts
Who changes what in their daily work, what each person loses or gains, before the solution is locked.
Prototype with future users
Card sorting, mock-ups, short tests: the solution takes in the field instead of being presented to it at the end.
Pilot on a limited scope
Business acceptance, named champions, feedback applied before rolling out.
Train, then measure usage
Training just before go-live, then tracking of real usage, not of accounts opened.
In a web agency: getting every role on the project to adopt AI
On a web project run in an agency, I deployed AI tools at every stage: SEO, lead generation, working prototypes, UX and help with functional specifications. Each stage had its own role, and each role its own way of working.
I did not start with the tools. I looked with each team at what was changing in their work, chose with them one specific task to hand over to AI, had it tested on a short scope, then set who reviews what before the result moves on. The tool changes; responsibility for the result stays with the people who do the job.
Tools I built to prepare adoption
- Brief EvaluatorA brief isn't a description: it's a set of decisions made. This tool checks the 12 dimensions a complete brief must cover and gives an immediate completeness score.
- AI Automation ROI CalculatorBefore recommending an AI tool to a client, the first question is always the same: is it actually worth the setup cost? This calculator produces an honest figure based on the real parameters of the task concerned.
- Project Risk MatrixThe risks that aren't written down are the ones that show up in crisis meetings. I built this tool to get project risks out of people's heads and onto a grid everyone can read and act on.
Frequently asked questions
What is change management?
Change management covers the actions that help the people affected by a project genuinely adopt a new organisation, tool or process: current-state review, impact analysis, communication, training, support for champions and usage tracking. It is managed as a workstream of the project, alongside the technical one.
When should change management start?
At scoping. Impact analysis and the involvement of future users shape the solution itself: started after development, they can only limit the damage. Training comes just before go-live, so teams apply what they learn straight away.
How do you measure the adoption of a new tool?
By tracking real usage rather than the number of accounts created: share of active users, frequency of use, tasks carried out in the tool rather than around it, and feedback from business champions. These indicators are set before launch, so they can be compared with the starting point.
How do you support the rollout of an AI tool?
With the same method, plus a governance step: deciding who reviews what the AI produces before it goes anywhere. AI changes the content of the work, not just the tool, which is why training covers critical review as much as usage.