BlogStrategy

Strategy

A Practical Agenda for an AI Discovery Workshop

Run an AI discovery workshop that produces a workflow map, a ranked problem list, and a testable project brief instead of a long collection of ideas.

An AI discovery workshop is useful when it turns different opinions into concrete choices. Invite the people who perform the work, the person accountable for its outcome, and someone who understands the relevant systems. Give participants a shared task to examine so discussion stays connected to actual operating conditions.

Bring evidence into the room

Ask participants to bring a few recent, appropriately shareable examples: a support request, a form, an internal article, or a report. Remove unnecessary personal information before the workshop. The aim is to understand the task, not to assemble a large dataset during a meeting.

Start by walking through one example from beginning to end. Note the tools used, handoffs, waiting periods, decisions, and rework. Let the person doing the task explain what makes a case difficult. Leadership priorities matter, but they should not replace the practical detail of how the work happens.

Explore options before choosing a solution

Describe the observed problems independently of any preferred technology. If information is difficult to find, possible responses might include better navigation, clearer ownership, structured filters, or AI search. Keeping these options visible helps the group avoid turning every inconvenience into a conversational interface.

For each promising idea, identify the input, proposed output, reviewing person, and fallback. In a hypothetical purchasing workflow, the first feature might summarize submitted requirements for a buyer to review. Automatically selecting a supplier would be a separate decision with different dependencies and consequences.

End with a brief that names unresolved questions

Produce a short project brief during the session: target user, current problem, proposed journey, source material, boundaries, and success evidence. Assign an owner to every unresolved dependency. “Confirm access to product documentation” is an action; “investigate data” is usually too vague to finish.

Reserve time to challenge the preferred idea. Ask which ordinary exception could make it impractical and which assumption is cheapest to test first. Then choose the next activity: a data review, a clickable prototype, or a limited technical experiment. A workshop should reduce ambiguity enough that someone can take that next step without another broad brainstorming session.