Where do we start with AI: tools for everyone, training, or one piece of work?
Four ways to start, and what each one leaves you with after a year.
The question
We run an established company of about a hundred people whose work is expertise-heavy — quotes, client files, technical reports. We have not really started with AI: a few employees use ChatGPT on their own, nothing is organised, and we cannot say where it would actually pay off. Our options are to give everyone a general-purpose AI assistant and let uses emerge, to train our teams first, to pick one piece of recurring work and rebuild it around AI, or to wait until the tools mature. Which starting point holds up, what does each one leave us with after a year, and if we pick one piece of work, how do we recognise the right one?
Counsel's position
Rebuild one carefully selected, recurring piece of work around AI to demonstrate value, build expertise, and mitigate risk.
Verdict
The verdict: Rebuild one carefully selected, recurring piece of work around AI to demonstrate value, build expertise, and mitigate risk.
How the criteria decide
2 of 3 criteria resolved on cited evidence.
| Criterion | Favours | Evidence |
|---|---|---|
| What visibly changes in the business after a year | Rebuild one piece of recurring work | Saving 30 minutes daily per 10,000 employees reclaims 600 FTEs The #1 driver of AI ROI is reimagining an end-to-end workflow. Not layering AI onto existing processes — rebuilding them from scratch with AI at the core Report personal AI productivity but only 37% see EBIT impact Eighty percent said AI had improved their own productivity. Yet only 37% said AI had contributed positively to their company’s EBIT |
| Time and attention it takes from our experienced people | Rebuild one piece of recurring work | Starting AI adoption with peripheral chores prevents expert resistance The goal is to use AI not to bypass the developer, but to relieve them of part of the peripheral work that currently slows down teams Agent-to-agent collaboration yields roughly output acceptance I stopped asking people to use AI to do their own jobs faster, and started asking them to hire and manage agents instead, like junior employees. |
| Cost, and what we keep if we stop | Not resolved |
Starting AI adoption with peripheral chores prevents expert resistance
Introducing AI through tasks employees dislike, such as retro-documentation, accelerates onboarding and prevents resistance from experts who value their primary work.
- Delphine Boudou (Pro à Pro): "The idea is to introduce AI through tasks developers don't like doing"
Managing AI agents like junior hires reached roughly 95% acceptance
Shifting employees from using AI as a personal tool to managing role-based agents accelerates development cycles and forces cross-functional collaboration.
Saving 30 minutes daily per 10,000 employees reclaims 600 FTEs
Rebuilding processes from scratch with AI at the core allows you to reskill employees for higher-value tasks, rather than just automating existing steps for minor time savings.
Report personal AI productivity but only 37% see EBIT impact
Giving everyone access to AI tools creates isolated productivity gains, but fails to improve company-level performance unless the underlying workflows are fundamentally redesigned.
Read another verdict
- Start with a small test, or take on the whole process at once?
- Our competitors advertise AI and we don't — match them, or hold the line?
- Our people already put client files into ChatGPT — ban it, frame it, or supply a tool?
- Our most experienced person retires in two years — how do we keep what they know?
- We can't hire the experienced people we need — automate, train up, or outsource?
- Slow our EU AI Act prep now the deadline's moved to 2027?
- Use AI to flatten middle management this year?
- Let an AI agent act on its own — or keep a human in the loop?