Choose tasks that are specific and safe to practise
Ask each participating role where repeat work involves drafting, extracting, comparing or restructuring information. Select a small task with a clear input and a result a knowledgeable colleague can assess. Avoid starting with high-stakes decisions, personal data or work whose correctness nobody in the room can judge.
Write down the current process and the friction you want to explore. A useful example might be turning approved notes into a first internal summary, not uploading a live customer file and asking the tool to handle everything. Define success as a reviewed draft or clearer method rather than a promised time saving.
- Realistic task with a defined owner
- Non-sensitive or sanitised example inputs
- Observable quality criteria
- Clear situations where AI should not be used
Confirm tools, accounts and data rules
Identify the AI service your organisation permits and the account type participants will use. Check access, licences, browser restrictions and relevant settings before the session. Do not assume that a consumer account and an organisation-managed account handle information in the same way; follow your own approved policy and vendor terms.
Prepare fictional or sanitised examples that preserve the shape of the task without exposing customers, employees, contracts or confidential strategy. Brief participants on prohibited inputs and on what to do if they are unsure. The facilitator also needs to know these limits so demonstrations do not contradict your rules.
- Approved service and account type
- Successful login before the session
- Written list of restricted information
- Prepared safe examples for each exercise
Brief participants, managers and reviewers
Tell participants the workshop purpose, schedule and what they should bring. Explain that the goal is not to accept the first answer but to practise giving context, checking output and deciding whether it is fit for use. Make accessibility or language needs known early enough to adapt materials.
Managers should identify who can approve future workflows and who remains accountable for work produced with AI assistance. Agree the review standard for facts, calculations, tone, confidentiality and completeness. After the session, schedule one small experiment and a review point instead of leaving each participant to invent a separate practice.
- Clear participant brief and expectations
- Relevant accessibility and language needs
- Named output reviewer or process owner
- Specific follow-up experiment and review date
Frequently asked questions
Frequently asked questions
How many tasks should we bring?
For a short session, one well-defined task is enough. A longer workshop can use several related exercises, but each needs safe inputs and someone capable of assessing the output.
Should everyone create an AI account before training?
Only if that account and tool are approved. Confirm licensing and access in advance. Some sessions can use a facilitator demonstration, but hands-on practice requires suitable participant access.
What should we avoid sharing with the trainer?
Do not send passwords, personal records, confidential customer information or restricted company material. Agree safe examples and necessary context during planning.