The most reliable way to train non-technical employees on AI is to teach small, role-based groups on their own real tasks, focus on a few durable habits rather than tool features, and follow up within weeks so the habits stick.
Non-technical people don’t need to understand how models work. They need to know what AI is good at in their job, how to set up a task well, and how to check the result.
Who this is for
Heads of people, COOs and team leads planning AI training for staff in operations, finance, investor relations, project management, administration or client service.
The short version
- Teach on real work, not generic examples. Recognition is what makes it click.
- Teach habits, not buttons. Interfaces change; good habits transfer across tools.
- Keep groups small and role-based, then follow up.
Principles that make training stick
Start from the work. Before any session, collect three to five tasks each role does every week: drafting a client update, summarizing a long document, preparing meeting notes, cleaning up a spreadsheet, answering a routine question. These become the training material, with sensitive details removed.
Keep it hands-on. People should spend most of the session using the tool, not watching slides. A useful rule of thumb is to keep presentation short and practice long.
Make it safe to be a beginner. Many experienced professionals are uncomfortable looking unskilled in front of colleagues. Small groups of peers, rather than mixed seniority, often help.
Say clearly what’s allowed. Cover approved tools and data rules at the start. Uncertainty about permission quietly kills adoption.
The four habits worth teaching
These habits apply whichever tool your firm uses.
- Give context. Say who the output is for, what it’s for, what good looks like, and include the relevant material. In our view, missing context is behind many weak results.
- Ask for structure. Request a specific format: a table, a short list of risks, a draft with headings, three options with trade-offs. Structure makes output easier to check and use.
- Iterate instead of restarting. Treat the first answer as a draft. Ask for changes: shorter, more formal, focus on section two, explain your reasoning.
- Check against the source. AI can be fluent and wrong. Teach people to verify facts, figures and quotes against the original document, and to own the final version.
A sample first session
A first session of about an hour, for one team, might run like this:
- Ten minutes: what AI is good and bad at in this team’s work, plus the data rules.
- Ten minutes: a live demonstration on one of the team’s real tasks, showing a weak prompt and then a strong one.
- Thirty minutes: everyone works on their own task from the prepared list, with help on hand.
- Ten minutes: each person shares one thing that worked and saves one prompt or checklist to reuse.
The goal is that everyone leaves with at least one task they will do differently that week.
After the first session
A single session creates interest. What happens next decides whether it becomes habit.
- Week one or two: a short check-in. What did people try? Where did they get stuck?
- Shared library: save prompts, checklists and good examples somewhere the team already looks.
- Champions: identify one or two people per team who enjoy this and give them time to help others.
- Second session: go deeper on harder tasks, such as multi-step work, reviewing long documents, or combining several sources.
- Leader example: ask managers to mention specifically how they used AI that week.
Adjusting by role
- Operations and administration: document handling, scheduling, routine correspondence, process checklists.
- Finance: explaining and drafting formulas, reconciliation support, commentary on variances, first drafts of reports, always checked against the numbers.
- Client-facing staff: meeting preparation, follow-up drafts, summarizing long threads.
- Senior leaders: board and meeting preparation, thinking partner for decisions, reviewing drafts. Often best trained separately and briefly.
Common mistakes
- Teaching features instead of tasks. People forget menus; they remember solving their own problem.
- One large, mixed session. Different roles need different examples, and seniority mixes can inhibit questions.
- Skipping verification. Teams that aren’t taught to check output either over-trust it or abandon it after one error.
- No follow-up. Without reinforcement, people tend to revert to familiar habits.
- Assuming younger staff don’t need it. Familiarity with chat tools isn’t the same as using them well at work.
When you do not need us
If you have someone internally who uses AI well, understands your teams’ work and has the time to prepare real tasks and run follow-ups, this approach can be run in-house. Use the session plan above. Outside help is most useful when nobody has that time, or when senior staff will respond better to an external trainer.
How Ruby helps
- AI training: role-specific, hands-on sessions built on your team’s real tasks, with follow-up and shared materials.
- AI strategy: deciding which teams and workflows to start with so training effort lands where it matters.
- AI engineering: turning the most repetitive trained workflows into systems when that makes sense.
Start with one practical hour of AI training for one team.