Start from work, not from technology
Ask teams: which tasks are repetitive, text-heavy and annoying? Emails to customers, offers, reports, meeting minutes, first-line support, searching internal documents. These are where general assistants deliver value within weeks.
Pick by value and risk
- High value, low risk (internal drafts, summaries): start now
- High value, higher risk (customer-facing, personal data): start with human review and clear rules
- Decisions about people (hiring, credit, evaluation): high risk under the AI Act — only with careful preparation
- Low value: skip, however exciting
A 30-60-90 day plan
- Days 1–30: approve one business tool, write a one-page policy, train a pilot group, choose three use cases
- Days 31–60: measure time saved and quality, collect the best prompts, fix what doesn’t work
- Days 61–90: roll out to more teams, add AI literacy training for everyone, decide on the next, bigger step
Your people are already using it
Surveys consistently show that many employees use AI at work without telling anyone — often with personal accounts. A clear “yes, with these tools and these rules” is safer than silence or a ban.
Measure honestly
Before the pilot, record how long the task takes and how many corrections it needs. After four weeks, measure again. Count review time too — a draft that needs 30 minutes of fixing isn’t a saving.