Generative AI Workshops for Business Teams

Build around work that already exists

Choose a recurring task such as synthesizing approved notes, drafting an internal brief or organizing a research comparison. Avoid starting with a large catalogue of features. The task should have a recognizable user, an output someone can judge and a clear boundary on what AI is allowed to do.

An enquiry to AISFY should describe the team’s work and current tool environment. ChatGPT, Claude, Gemini and Microsoft Copilot may be relevant depending on access and task requirements. Product features and data settings vary by plan and organization configuration; they must be checked before preparing exercises.

Use a five-part task brief

A brief makes the task inspectable by another person. If the team cannot describe a good result before using AI, it will struggle to evaluate a plausible result afterwards. Keep the brief and reviewed output together so the exercise can be repeated.

• Purpose: who needs the output and what decision or action it supports.
• Inputs: which approved sources can be used and which information must be excluded.
• Output: the required structure, detail and audience.
• Quality: the facts, calculations, tone and omissions a reviewer must check.
• Boundary: what the tool must not decide, publish, send or change.

Example exercise: an internal decision brief

Give participants a fictional project update containing three options, stated costs, dependencies and missing information. Ask them to produce a one-page brief for a manager. The tool may organize the material and suggest questions; it may not invent missing figures or choose an option on behalf of the manager.

The reviewer marks each statement as supported, unclear or unsupported. Participants then revise their task brief and repeat the exercise. Discuss whether the improved output came from better source material, a clearer task or more careful human review. This makes the learning transferable to a different tool.

Match exercises to the team

Prepare access and safe materials

Confirm accounts, licences, administrator settings and connectivity before the session. Use sanitized or fictional examples by default unless the organization has approved a different arrangement. Do not ask participants to bypass access restrictions or use personal accounts to solve an access problem.

Provide an alternative exercise for participants who cannot use the selected tool. Pair work can support discussion, but distinguish observing a demonstration from completing an independent task. Accessibility, language and digital-confidence needs should shape the facilitation plan.

What counts as a useful output?

A useful output includes the task brief, the source used, the reviewed result and the corrections required. Record failures as well as successful attempts. If the tool invents facts or cannot work within the data boundary, that is information for deciding whether the task is suitable, not an inconvenience to hide.

A session can establish an initial learning signal. Workplace application still requires management support, approved access and a suitable task after the workshop. Any claim of time saved must include review and rework time as well as the initial draft.

Choose the next step deliberately

At the end, select a modest practice task and assign a reviewer. Decide what evidence will be reviewed and when. If the main obstacle is unclear policy, fix that before expanding tool use. If the obstacle is inconsistent output, focus the next practice on evaluation rather than adding more features.

Frequently asked questions