AI Workflow Workshop: A Practical Agenda for Your Team
An AI workflow workshop brings a team together for a few hours to turn one real process into a tested, agreed workflow with a review checklist. Here is a practical agenda and what to do after.
Updated 6 min read
An AI workflow workshop brings a team together to turn one real process into a tested, agreed AI workflow with a review checklist. The goal is not to discuss AI in general; it is to leave with one workflow the team can run next week and a record of the rules and lessons that came out of building it. Tools like fenbs keep those agreed rules and the follow-up tasks in one place, so the workshop produces something the team builds on rather than a document that sits in a folder.
What a workshop should produce
A workshop that ends without a concrete output was a meeting, not a workshop. Define the deliverables before you send the invite. A useful set of deliverables for a half-day session includes:
- One workflow defined: inputs, steps, owners, checkpoints, and the out-of-scope list.
- A review checklist: the criteria a person applies before the AI output moves forward.
- A set of agreed rules: decisions the team made during the session that apply to every future run.
- A list of follow-up tasks: things to test, tools to evaluate, or decisions still to be made.
- A lessons log: what went wrong during the in-session test and what the team learned from it.
If the team cannot agree on a single workflow to improve, the workshop is not ready to run. Pick the workflow first, before the session, and confirm it with the participants.
The pre-work: choosing a candidate process
Send participants one question two to three days before the session: "Which one process in your regular work takes the most time but produces an output you could check in five minutes?" The answers to that question are the workshop's candidates.
A good candidate process for a first workshop has a clear output the team can evaluate, runs at least weekly, and does not depend on information that is confidential or legally protected in a way that prevents sharing with an AI. A process that meets those three criteria is ready to workshop.
Send one example of the process input and one example of a good output to participants before the session. Seeing a real example focuses the discussion faster than a description.
A workshop agenda
The agenda below suits a half-day session of three to six people. Adjust the timings for your team size and the complexity of the process.
| Time | Activity | Output |
|---|---|---|
| 0:00 to 0:15 | Frame the goal: one workflow, ready to run by next week. Review the candidate process and the example input and output. | Shared understanding of the task. |
| 0:15 to 0:45 | Map the current steps: what does a person actually do, in what order? Include the checks and decisions, not just the actions. | A whiteboard or shared doc with the current steps listed. |
| 0:45 to 1:15 | Identify the AI steps: which steps could an AI handle if given the right input and constraints? Assign each step to person or AI. | A draft step list with owners. |
| 1:15 to 1:30 | Break. | |
| 1:30 to 2:00 | Write the rules: what must always be true about the output? What is the AI allowed to decide alone and what needs a person? | A draft rules list. |
| 2:00 to 2:30 | Run the workflow live: use the example input, run the AI steps, and apply the review checklist. Note every problem. | A test run result and a list of problems found. |
| 2:30 to 2:50 | Fix the problems and restate the rules: update the workflow definition based on what the test found. | An updated workflow definition. |
| 2:50 to 3:00 | Assign follow-up tasks: who does what before the first real run? | A task list with owners. |
The live test in the 2:00 to 2:30 slot is the most valuable part. A workflow that looks reasonable on a whiteboard often reveals a missing input or an ambiguous rule when you run it on a real example.
Running the workshop session
One person facilitates and one person records. The facilitator keeps the group focused on the current step; the recorder writes down rules, problems, and decisions as they come up. Separate those two roles before the session starts.
When the group disagrees on whether a step belongs to a person or an AI, apply one test: if the AI gets this step wrong and nobody checks it before it moves forward, what is the worst-case outcome? Steps where the worst case is significant stay with a person. Steps where the worst case is correctable in the next review pass are AI candidates.
For education teams and similar groups running a workshop for the first time, keep the scope of the first workshop to one process and one team. Expanding to a second process in the same session usually produces two incomplete definitions instead of one complete one.
The follow-up: agreed rules and next tasks
Send a summary within 24 hours. The summary should include the agreed workflow definition, the rules list, the lessons from the test run, and the follow-up tasks with their owners. A summary that arrives a week later competes with the team's memory of what was decided.
The agile-workflow guide covers how to treat the workflow definition as a living document that the team revises after each run. The first version that comes out of a workshop is a starting point, not a final answer. The ai-agent-approval-workflow guide covers how to build the approval step so the team can see which AI outputs are passing review and which are not.
Turn the workshop output into a shared record
fenbs keeps the agreed rules as Decisions and rules, the follow-up tasks on the board, and the lessons from the test run in one record the whole team can read. Add everything from the workshop summary before the first real run. Connected assistants read the rules and lessons before they start, so the output of the workshop actually reaches the work.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
How to do this with fenbs
After the workshop, add the workflow as a project and transfer the output from the session summary. Record the rules and the follow-up tasks before the session summary is the last place the team looks.
- Create a free account. Add a project for the workflow and paste the workflow definition as the project description.
- Add the agreed rules as Decisions and rules. Record who decided each one and the reason from the workshop discussion.
- Add the follow-up tasks to the board with owners and the agreed next step for each.
- Record the lessons from the test run. What went wrong, what the team learned, and what to do differently on the first real run.
- Connect your AI app from Integrations. Ask it to read the project context and rules before the first real run.