70% AI adoption, starting from zero

The problem

• The tools were available, but people did not know what they were allowed to do or how to start.
• The environment sat behind a corporate firewall, so off-the-shelf training that pulled from the internet did not work.

What was done

1. Built a curriculum on prompting, how AI models use context, and fitting AI into real daily work, packaged to run entirely inside the firewall.
2. Taught it directly, then recruited peer advocates on each team to keep it going.
3. Built verification steps into every procedure, so people check an AI's output before they rely on it.
4. Designed the setup so the AI loads only the instructions a task needs, which keeps usage costs down as adoption grows.

The result

• About 70% adoption, starting from no prior AI experience.
• Verification steps built into every procedure from day one.
• Where it was put to work: AI for everyday business tasks, and AI that writes SQL database queries so non-technical users can get answers from the data themselves.

What it shows

AI adoption is a change-management problem more than a technology problem. The tools were already there; the missing pieces were confidence, guardrails and people who help each other.