AI training for staff: a rollout plan for SMEs
Train the whole team on its own work, not on a generic syllabus. Start with a short entry test and one named task per person, run 6–8 two-hour sessions over 3–4 weeks, finish with a demo day and a mirror test, and give leadership a before-and-after report. Then turn the best results into the first automations.
Why doesn’t most AI training stick?
Most teams already use ChatGPT, Claude or Copilot on their own. Each person saves a little time, but nothing joins up: answers sit in private chats, knowledge stays in people’s heads, and the processes look exactly as they did before. A one-day workshop rarely changes this, because people go back to the same work with the same habits the next morning.
What changes it is time and relevance: several sessions with practice in between, on tasks the person actually does, with something measurable at the end.
What does a four-week rollout look like?
Week 0: discovery and an entry test
Each participant names one task they want to hand to AI and answers a short test on what they know and use today. The programme is finalised from those answers, so a sales team gets proposals from call transcripts and finance gets reporting scripts for its own sheets.
Weeks 1–2: foundations on real work
Prompt structure (role, context, task, format), ChatGPT and Claude as workspaces, custom assistants per department, and a knowledge assistant on your own contracts, policies and product documents. Every session ends with homework on the participant’s named task.
Weeks 3–4: data, presentations and automation
Reporting and small scripts for the sheets you already use; presentations to your brand book; a first automation on a real scenario with n8n or MCP. A session on what must never go into a model.
The last session: demo day and the mirror test
Each person shows what they built. The exit test mirrors the entry test, and leadership receives a short report: how many people now use AI daily, which tasks moved, and time saved per task.
Who should you train first?
Start with the departments most willing to try — usually marketing and HR — not the ones with the biggest problem. A visible early win makes the rest of the company ask for it, which is faster than persuading a sceptical team first. Run a short separate track for the leadership team so that the people who set priorities have used the tools themselves.
What should you measure?
- People using AI at work at least weekly, before and after.
- Tasks moved to AI, and time per task before and after, for three or four named tasks.
- Entry and exit test scores.
- Assistants and automations built by the team itself.
At ImLink, eight sessions over four weeks took the team from one person using AI to fourteen, and the team built nine agents of its own. Read the case study.
What comes after the training?
Training doubles as discovery. By the last session you know which processes people automated on their own, which they asked for help with, and where the data lives. That is the input for an audit or straight for the first automation pilots — and, once calls and documents are in order, for a company brain.
Questions
How many sessions does a team need?
Six for a compact start, eight for full tool coverage plus a final test, up to thirteen if you add integrations, n8n and MCP or industry-specific work.
Online or in person?
Online works well for two-hour sessions twice a week; it is easier to keep the rhythm and everything is recorded. In-person days suit kick-offs and demo days.
Which tool should we train on?
The one your company will standardise on. If that is not decided, train on two — usually Claude or ChatGPT plus Copilot for Microsoft-centred teams — and choose after the first sessions.
How we write these guides: figures about our own work come from before-and-after measurements on client projects (names withheld where under NDA); market figures come from the sources listed, with the date we checked them. Prices exclude VAT.
