TL;DR: A system without training is a Segway — beautifully engineered, nobody rides it. Whether you migrated or are onboarding — training is what makes people use it.
The hard part
Giving people the system is step one.
Teaching them to use it is step two.
Step two matters more. A clean vault, a good folder structure, and an approved AI tool mean nothing if people keep doing this:
- Asking the same person for context.
- Saving files in the old place.
- Pasting confidential data into random chatbots.
- Trusting AI output without checking it.
- Avoiding the new system because it feels slower.
Adoption does not happen overnight. It is repeated practice with consistent training.
What people need to learn
Start with the fundamentals — Markdown syntax and, if your company uses it, Git basics. Point your team to the Markdown Essentials guide and the Git Essentials for Company Files as prerequisites.
Do not train people on “AI” in general: train them on their work.
Every team needs five basics.
1. What AI is good at
Show real work, not demos.
AI is useful for:
- Finding the right internal document.
- Summarizing meeting notes.
- Turning messy notes into a checklist.
- Drafting a first version of an email, report, or SOP.
- Comparing two versions of a policy.
- Extracting action items.
- Rewriting content in the company tone.
The simple message:
AI removes first-draft friction. It does not remove responsibility.
2. What AI is bad at
Say this clearly.
- AI can be wrong.
- AI can invent facts.
- AI can miss context.
- AI can sound confident while being useless.
- AI should not make final decisions.
- AI should not touch sensitive data unless the workflow is approved.
The rule everyone should remember:
AI drafts. Humans decide.
That sentence is the seatbelt.
3. How to give context
Context is the information you give AI so it can answer properly.
Bad prompt:
Write the process.
Better prompt:
Use this SOP, this customer note, and this policy. Draft a 7-step support process. Keep it short. If something is missing, flag it instead of inventing it.
Teach people to include:
- The task.
- The source documents.
- The audience.
- The expected output.
- The constraints.
- What AI should do when information is missing.
No context, no quality. Garbage in, garbage eloquently out.
4. Who owns the output
The person using AI owns the result.
That means:
- Read before sending.
- Check facts.
- Check sources.
- Check tone.
- Never blame the tool for a bad output.
If you would not sign your name under it, do not send it.
5. What is safe
Start strict, loosen later.
Basic rules:
- Use the approved AI tool for company work.
- Do not paste restricted data unless approved.
- Do not use free personal chatbots for confidential work.
- Do not let AI send external messages without review.
- Do not let AI delete files unless absolutely sure.
- Do not let AI change permissions.
- Do not let AI make hiring, legal, finance, or disciplinary decisions.
- Report mistakes and near misses without fear.
The brakes must be in place.
The initial training session
Run one 60- to 90-minute workshop per team.
Keep it practical:
- Explain the system. Where is the source of truth? Which tool is approved? What changed from the old way of working?
- Explain the rules. What data is allowed? What is forbidden? Who checks the output?
- Show three real examples. Find an SOP. Summarize a meeting. Draft a handover note.
- Make everyone do one task. Each person should leave with something useful from their own work.
- Ask what felt wrong. Resistance is feedback. Silence is expensive.
No toy examples. Toy examples create toy adoption.
After the workshop
For the first month:
- Run weekly office hours.
- Collect good prompts in the vault.
- Share before/after examples.
- Fix confusing folders quickly.
- Ask skeptics what still feels slower.
- Give champions permission to help teammates.
- Review AI mistakes without blame.
The system improves when people trust it enough to complain honestly.
Adoption checklist
Use this after each team rollout.
- Team knows where the new source of truth is.
- Team knows which AI tool is approved.
- Team knows what data cannot go into AI.
- Team practiced on real work.
- Team knows humans own AI output.
- Team has a place to ask questions.
- One champion is helping the team.
- At least one workflow saves time this week.
If the last item is false, fix that first.
Final recommendation
The system makes knowledge readable.
Training makes it usable.
Do not aim for excitement. Aim for a boring sentence from the team:
This saved me time. I will use it again. That is success.
What’s next
Training framework is set. Now teach the basics: Markdown Essentials — the writing format your team will use every day.
Your data. Your rules. Let’s write it that way. By Charles Henri Gayot.