The gap is common. Teams experiment, learn and work differently. The organization keeps running as before. Microsoft’s latest Work Trend Index helps look this gap in the face.
The paradox of employees who are equipped but blocked
The 2026 Work Trend Index, published by Microsoft on May 5, 2026, draws on a survey of 20,000 AI users in ten countries. In France, 49% of users say they produce work they could not have done a year earlier. But only 20% believe their leadership is clearly and consistently aligned on AI priorities.
A note on method: these data are mainly self-reported. They describe perceptions, not measured results. The gap is still telling. Individuals are moving forward, and the organization’s direction looks unclear to them.
Why access to Copilot is not enough
Buying licenses and opening the platform does not mean the organization allows new ways of working. An employee can know how to use Copilot without having the time, the legitimacy or the manager’s agreement to change how they work.
Microsoft reports that organizational factors, notably culture, managerial support and talent management, account for 67% of the reported impact of AI, against 32% for individual factors. I read this figure with caution. It concerns reported impact. It does not prove that the human factor alone explains the success of a program: technology, data and security matter too.
It does point to a frequent blind spot. We steer the deployment, and rarely the conditions of use.
Old rules of work block new ways of working
In many cases, objectives, performance indicators and approval workflows were built for the old ways of working. An indicator that rewards volume processed does not encourage anyone to check a summary produced with AI. An approval process designed for a hand-written document can cancel out the time saved.
The employee then faces a simple choice: follow the rules or change their practices. Most follow the rules.
Deloitte, in The State of AI in the Enterprise (January 2026), notes that only 34% of the organizations surveyed say they are deeply reinventing their processes, products or business models. I read this figure as follows: many organizations add AI to ways of working that have not changed.
What managers must make possible
Frontline management is the tipping point. It turns the leadership’s ambition into daily rules. Three levers seem decisive to me:
- Show the uses: a manager who uses AI in front of the team is more convincing than a framing memo.
- Set quality criteria: what is a good result produced with AI, and who checks it?
- Create a right to supervised experimentation: a clear scope, authorized data, dedicated time, and the freedom to say what did not work.
Without this, uses remain individual and discreet. Nobody talks about them, so nobody learns from them.
Train on real tasks rather than on features
Deloitte finds that the skills gap is the main obstacle to integrating AI into workflows. Training therefore matters. But I distinguish three workstreams, which are often confused:
- Training: learning to use the tool
- Role evolution: deciding what each person now does, and no longer does
- Workflow redesign: redrawing the steps, approvals and handovers
Generic training covers the first. It does not settle the other two. A workshop that starts from a real task, for example reviewing a contract or preparing meeting minutes, forces the questions of role and flow. That is where usage becomes lasting.
Signals of an organization that is not prepared
A few signals reveal it without a heavy survey:
- Teams ask for permission to try rather than making proposals
- Managers cannot describe a good AI-assisted result
- This year’s objectives have not changed since the deployment
- AI-assisted work goes through the same approvals, or more
- Accounts are active, but nobody knows which tasks have changed
- Leadership supports AI in speeches, without deciding what teams can stop doing
Putting real usage back at the center of steering
The number of active accounts or prompts shows that the tool is open. It does not show that the work has changed. Real adoption is measured in practices: which tasks are now done with AI, with what quality, in what time, and which processes were modified to allow it.
I recommend tracking these indicators with managers, not only with IT. And linking each use case to a business result the team recognizes as its own.
Employees are often ready. The useful question for a sponsor becomes: what is my organization preventing them from doing?
Is your AI program moving slower than planned? Let’s talk about identifying what is blocking it, realigning stakeholders and getting adoption moving again.