Start with a shared safe-use baseline for everyone: which tools are approved, which data may go in, and who checks the output. Then run role cohorts where managers practise those rules on their own work. A one-off workshop raises awareness, but habits change only when training sits inside the job people already do.
#Key takeaways
- Your managers are already using AI, so training should start from what they do today, not from zero.
- A short safe-use baseline for all staff covers approved tools, data rules and human review.
- Role cohorts then apply that baseline to real workflows, which is where habits form.
- Measure usage and incidents before and after, so HR can show what changed.
- Banning tools without offering a better approved option pushes use out of sight.
#How common is it for staff to bring their own AI tools?
Very common. The Microsoft and LinkedIn Work Trend Index, may 2024 reported that 75% of knowledge workers use AI at work and 78% of AI users bring their own tools, a pattern it calls bring your own AI. Only 39% of AI users said their company had given them AI training.
So the managers pasting reports into ChatGPT are not unusual. They found a tool that helps and nobody told them the rules. The risk is real, because customer data, salaries or contracts may leave your systems, but the instinct to use AI is one to channel, not punish.
#What should the safe-use baseline cover?
Keep it short enough that every employee finishes it, and concrete enough that they remember it. Stennir builds it around four points:
- Which tools are approved, and which accounts to use them through.
- Which data may never go in: personal data, client files, salaries, anything under contract.
- Who checks AI output before it reaches a customer, a regulator or a payslip.
- How to report a mistake, such as pasting the wrong file, without fear of blame.
The baseline also explains, in plain words, why prompt injection and leaked data matter. The OWASP top 10 for LLM applications lists prompt injection as its first risk, which is a useful reference for it and security teams setting the rules.
#What works beyond a one-off workshop?
Role cohorts. After the baseline, group people by role: finance managers together, HR partners together, operations leads together. Each cohort works on one real workflow from its own week, using approved tools and its own documents. The trainer has done that job and shows the prompts and checks that actually hold up.
Buyers who sat through long generic courses tend to say the same thing: they did not like the training. Short sessions on real tasks, with confident users paired alongside hesitant ones, do better. The cohort ends with a shared set of prompts and checks that the team keeps using.
We are
Stennir gives every employee one safe-use baseline, then trains role cohorts on their own workflows with approved tools.
We aren't
Stennir is not a policy writer who emails a PDF, and not a vendor pushing a single AI product.
#How does HR show the training worked?
Measure before and after. Record which approved tools each role uses and how often, plus any data incidents reported, before the cohorts start. Measure again afterwards. The AI training service ends each programme with an adoption report built on those numbers, so HR and L&D can show the change instead of counting attendance.
#What should an HR or L&D lead do next?
Ask a few managers what they use AI for today; the answers tell you which workflows to train first. Then take the free readiness assessment or book a 30-minute call. For the operations angle, see AI for operations, and for low Copilot usage, read why Copilot licences go unused.
#Common questions
Do we have to ban ChatGPT first?
Not necessarily. A ban without an approved alternative tends to move use onto personal phones, where you see nothing. A better order is to name the approved tools, set the data rules, and train people on both. Some companies then allow ChatGPT through a business account with clear limits on data.
What do you need from us to start?
Stennir needs your current AI or acceptable-use policy if one exists, a list of approved tools, and a contact in HR or l&d. For the role cohorts, we need one person per role who can describe a real workflow and share sample documents that are safe to use in training.
Does the training work in Arabic and French?
Yes for Arabic and English, which Stennir supports fully. French and other languages are available on request. The baseline and cohort materials are written in the language your staff work in, and examples use your own documents, so people practise on the kind of text they handle every day.
Where does our data go during training?
It stays in your accounts. Cohorts use your approved tools under your own tenancy, and Stennir does not copy training documents into its own systems. Where personal data is involved, the examples are redacted or replaced with realistic samples before the session, so nobody practises on real customer records.
