Last Thursday, a manager tells me: "I'm behind on AI." He's testing ChatGPT, Copilot, Gemini, Claude. So is his team. But nobody can say if it's actually creating value.
He's not behind. He's caught in the speech trap.
88% of companies already use AI. Only 32% have truly integrated it into their ways of working. There's plenty of AI in the slides. Very little in the workflows.
The false problem: "we need to convince people of AI's potential"
The classic boardroom. PowerPoint slides on "AI's transformative potential". McKinsey studies cited. 30% productivity promise.
Three months later, results inventory. A few licenses distributed. Zero business metrics moved. And 3 out of 4 employees using ChatGPT in shadow AI on their personal phones, because it's faster than the official process.
Speech convinces the ear. It doesn't enroll the working hand.
The real problem: trust doesn't precede usage, it follows from it
In the field, the same scene repeats. A team I supported was using 4 AI tools in parallel. Lots of testing, zero clarity on the value produced. We moved to 1 single tool, truly integrated into their daily rituals.
+40% measurable gains in 3 months.
Not because we found the best tool. Because we stopped stacking and redesigned a workflow around concrete usage.
75% Shadow AI isn't primarily a security problem. It's a signal — your employees seek meaning and find individual workarounds because nobody offered them an actionable framework. The innovation energy is there, waiting to be channeled instead of stifled.
The shift: optimization vs reinvention
McKinsey looked at the 6% of "high performer" organizations actually capturing AI value. The pattern:
→ They don't settle for optimizing, they reinvent
→ They redesign workflows end to end, not task by task
→ They change roles, not just tools
→ They align leadership, processes and products toward an AI-Native model
AI isn't here to do "the same thing, faster". It's here to enable what was impossible yesterday.
Put differently, the strategic question changes nature:
Wrong question — "How can AI help on this process?" → you get an assistant.
Right question — "What work disappears if AI does it well?" → you redesign the entire project.
Not the same question. The first looks for a tool. The second changes who does what, redefines what we expect from a human in the team, shifts responsibilities.
Convincing through usage: standardized small wins
You don't reassure with slides. You reassure by showing how AI concretely improves a quote, a client follow-up, a summary note, a schedule. Small win, then another, then a standard.
4 levers I systematically apply on missions:
1 — Start from the work gesture, not the tech. 90-minute workshop with a team. Start from their time-eating irritant. Enter AI through their problem, not through a generic demo.
2 — Normalize the popular tools, don't impose top-down picks. An honest Shadow AI audit often reveals the truly useful tools. Better to secure ChatGPT in enterprise version than impose a Copilot nobody opens.
3 — Measure the gesture, not the license. "1,200 licenses activated" says nothing. "The standard quote went from 45 min to 12 min" says everything. ROI is measured as close as possible to the work.
4 — Capitalize on what works. Each small win becomes a documented pattern, shared across teams. Otherwise each manager reinvents alone, and the methodological capital stays at zero.
Monday morning, the work-gesture test
To know if your AI is in the slides or in the workflows:
- Ask 5 random employees to describe a concrete work gesture they do differently in the past 3 months thanks to AI. If the answer is fuzzy, AI isn't yet in the work.
- List the AI tools officially deployed and look at their real weekly usage rate — not the number of licenses. If the gap exceeds 50%, you have a usage design problem.
- Ask managers what changed in their decisions since AI arrived. If the answer is "nothing", AI is just a formatting assistant.
- Map Shadow AI without punishing. It's your best indicator of usages that genuinely create value — and of those your official deployment missed.
AI doesn't boost your current operation. It forces you to create a new one. Organizations that stay on the surface will lose their operational advantage to those that descend to the work-gesture level.
In your organization, is AI still a tool, or already a change in the way you work?