AI Workshops for Companies: Format, Duration and What You Actually Get
A workshop is not a talk with exercises. It only pays off when participants work on their own tasks and leave with something that survives inside the organisation. This article describes a structure that has proven itself in practice — including the points where workshops regularly fail.
Preparation: no use cases, no workshop
Two weeks before the date we ask every participant for two or three recurring, time-consuming tasks from their daily work. That list is the raw material of the day. Without it you get a tool tour that is forgotten within three weeks.
In parallel we clarify the boundaries: which systems are in use, which data may be used, and whether an internal policy for AI tools exists. If there is none, a first draft is often the most valuable outcome of the day.
Agenda of a full workshop day
- Orientation (45 min) — what language models do, what they do not do, and why they are sometimes convincingly wrong.
- Ground rules (30 min) — which data may go in and which may not: customer data, HR data, contracts, health data. FADP and GDPR in practical rather than legal terms.
- Craft (90 min) — participants work on the tasks they brought: write instructions, check results, refine.
- Automation (90 min) — from a single task to a repeatable process: templates, interfaces, where a real integration pays off and where it does not.
- Assess and prioritise (60 min) — all ideas on an effort-benefit matrix, then the first three steps with owners and dates.
Group size, duration, composition
- 8 to 16 people — fewer and you lose the breadth of examples, more and not everyone gets to work.
- Half day or full day — a half day covers orientation, ground rules and first attempts. Automation needs a full day.
- Mixed roles — business, IT and one person from management. Without decision-making authority in the room the results stay non-binding.
- Own devices — people work on the systems they will actually have available afterwards.
What exists at the end of the day
- A prioritised list of use cases with an effort and benefit assessment.
- A draft internal usage policy: permitted data, prohibited data, review duties, responsibilities.
- A collection of tested working instructions for the most frequent tasks.
- Three concrete next steps with names and dates.
Everything else — satisfaction scores, nice photographs — is pleasant but not an outcome.
The four most common reasons an AI workshop fizzles out
- No mandate from the top. If management delegates the topic and never asks about it again, nothing happens.
- No tools afterwards. People who are enthusiastic on Tuesday and have no access on Wednesday lose interest.
- No rules. Without clear data protection guidance you get either paralysis or uncontrolled growth — both expensive.
- First steps that are too large. An initiative with a six-month horizon dies before the first visible result. Two weeks is the right size for step one.
Frequently asked questions
How long does an AI workshop take?
Half a day (three to four hours) for fundamentals and first hands-on use, a full day if automation and a usage policy should also come out of it.
Do participants need prior knowledge?
No. The only requirement is that each person brings two or three of their own recurring tasks.
May real company data be used during the workshop?
Only after clarification. We normally work with anonymised or non-critical data until the internal policy is in place and the environment has been reviewed.
Can the workshop take place at our offices?
Yes, anywhere in Switzerland and Liechtenstein. Remote delivery is possible, but on-site is noticeably more effective for the working phases.
Planning a talk or a workshop?
Send us the date, the audience and your time slot — we reply within two working days with a topic proposal.