To run an AI workshop on your team's real work, collect each person's tasks and documents beforehand and agree the data rules. Then spend most of the session hands-on. End by ranking three to five problems where AI could help, each with an owner and a first step. Write up what worked within a week.
Key takeaways
- A good AI workshop is mostly hands-on, using the team's own tasks, documents and tools.
- Preparation decides the session: tasks collected, sample documents cleaned, accounts working and rules agreed.
- Include one task where AI is weak, so people practise checking output as well as producing it.
- According to Pew Research Center, 52% of US workers are worried about AI's future impact on the workplace.
- The output is a ranked list of problems with owners and first steps, written up within a week.
What is an AI workshop on real work?
An AI workshop on real work is a hands-on session where one team uses AI on tasks from its own week, with its own documents and tools. It replaces the usual demo or lecture. People leave knowing where AI helps in their job, where it does not, and what they will try next.
The difference from a typical session is the material. A generic workshop uses made-up examples, so people must translate everything back into their own job afterwards, and most never do. A real-work workshop skips that step. If a finance officer spends the session reconciling a real supplier statement, the method is already theirs.
It is also a single team at a time. Sales, finance and customer success have different AI problems, so mixing them dilutes the session for everyone.
How should you prepare for an AI workshop?
Prepare five things: a list of tasks from each attendee, sample documents with personal information removed, working business accounts on the team's AI tool, the data and approval rules agreed by leaders, and one strong example from the team's best person. Preparation takes longer than the session and decides whether it works.
- Tasks. Ask each person for three tasks from a normal week, with how long each takes.
- Documents. Collect real examples, with names, contact details and financials removed.
- Accounts. Check everyone can sign in to the business plan before the day.
- Rules. Confirm what data may go into the tool and what needs approval.
- Example. Ask the person the team trusts most to show how they handle one task.
The rules come from leadership, not the facilitator. If nobody has decided what company information may go into which AI, settle that first. The guide to the decisions leaders make covers it.
Who should be in the room?
Invite one team, small enough that everyone works hands-on: the people who do the work, their manager, the team's AI champion and someone who can answer questions on the data rules. Keep senior leaders who are not part of the team out, so people feel free to admit what they do not know.
Expect worry as well as interest. According to Pew Research Center's survey of 5,273 US workers, 52% say they are worried about the future impact of AI use in the workplace. Only 36% say they feel hopeful. Some of your team will arrive quietly wondering whether the session is a prelude to job cuts.
Name the worry at the start rather than hoping nobody raises it. Say what the business has decided, and what it has not. The guide to talking about AI and jobs suggests the words.
What does an AI workshop agenda look like?
A practical agenda runs in six blocks: the rules, a real example from a colleague, hands-on practice on each person's tasks, a task where AI is weak, sorting and ranking the opportunities, and commitments. Spend at least half the time hands-on. Keep the talking short, because confidence comes from doing.
| Block | What happens | Output |
|---|---|---|
| Rules | Data, accounts and approval, on real examples | Everyone knows the limits |
| Show | A colleague demonstrates how they do one task | A method worth copying |
| Try | Each person works on their own tasks | First results on real work |
| Check | Everyone tries a task where AI goes wrong | Practice at spotting errors |
| Rank | Group sorts and ranks opportunities | Three to five priorities |
| Commit | Each priority gets an owner and a first step | Next week's work |
The check block matters most. In a BCG experiment, even participants warned that the tool could give wrong answers "did not challenge its output". Practising on a task where AI fails, in a safe room, builds the habit of checking. The sorting exercise in the guide to AI training that works suits the rank block.
How do you handle sceptics and enthusiasts?
Give sceptics the hardest task and let the result speak, whichever way it goes. Ask enthusiasts to help a neighbour rather than show off. Treat every objection as useful information about where AI is risky. A workshop that converts nobody but produces honest rules has still done its job.
Sceptics are often right about something specific. A senior adviser who says "it gets our pricing wrong" has found a real risk. Write it down as a rule: pricing always needs a person's check. That turns resistance into the kind of judgement the team needs.
Enthusiasts can unsettle the room by racing ahead. Pair them with someone less confident, which builds the peer support a champion later carries on.
Blocking AI rarely ends its use. According to BCG's 2025 AI at Work survey, when employees lack the AI tools they need, more than half say they will find alternatives and use them anyway. So treat the workshop as a chance to bring that use into the open, on approved accounts.
What should happen after the AI workshop?
Within a week, send the team a short write-up: the ranked priorities, each owner and first step, the rules agreed and the methods that worked. Put those methods into shared files the team's AI reads. Then check progress at a follow-up session and at each monthly clinic.
The write-up is where most workshops fail. Without it, the energy fades and the ranked list sits on a flip chart photo nobody opens. Keep it to one page so people actually read it.
The methods matter more than the list. When a colleague's way of handling a task is written down and checked with them, everyone's AI can follow it, including new starters who were never in the room. That is how one session keeps paying back.
Hours count too. According to BCG, regular use is far higher among employees who receive at least five hours of training, so plan follow-up sessions rather than relying on one day.
Skein offers adoption as one of three ways in, alongside strategy and implementation, and a business can begin at any one; the AI facilitation page covers how teams learn on their own work. That suits a team that has licences but no shared way of using them. A team with no agreed rules on data should settle those with leadership first.
Frequently asked questions
How long should an AI workshop for a team be?
A half day to a full day suits a hands-on workshop, with at least half the time spent working on real tasks. Shorter sessions rarely leave room for practice and ranking. Follow it with shorter sessions in the weeks after, so people bring back questions from using AI on real work.
What should employees bring to an AI workshop?
Each person should bring three tasks from a normal week, with real examples of the documents involved and any personal or financial details removed. They should also have a working sign-in to the business AI tool. That way they practise on work they will do again the following week.
Who should run an AI workshop?
Someone who understands both AI and the team's work, whether an internal lead or an outside facilitator. They should keep the talking short, keep everyone hands-on and handle worries openly. The team's own best performer should demonstrate at least one task, because colleagues trust a peer's method.
Should different teams do AI workshops together?
Usually not. Sales, finance, marketing and customer success have different tasks and different risks, so a mixed session ends up generic. Run one workshop per team on its own work. A short company-wide kickoff beforehand can cover the shared rules and what is changing for everyone.
What is the output of an AI workshop?
A one-page write-up: three to five ranked problems where AI could help, each with an owner and a first step, plus the data and approval rules agreed and the methods that worked. Those methods should then be written into shared files so everyone's AI can follow them.
Related reading
- Pew Research Center. "U.S. Workers Are More Worried Than Hopeful About Future AI Use in the Workplace." Pew Research Center, 25 February 2025. https://www.pewresearch.org/social-trends/2025/02/25/u-s-workers-are-more-worried-than-hopeful-about-future-ai-use-in-the-workplace/
- Candelon, François, et al. "How People Can Create, and Destroy, Value with Generative AI." BCG, 21 September 2023. https://www.bcg.com/publications/2023/how-people-create-and-destroy-value-with-gen-ai
- BCG. "AI at Work 2025: Momentum Builds, but Gaps Remain." BCG, 26 June 2025. https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain