An example AI workshop agenda built around actual team tasks

Suppose a team needs to turn meeting notes into a handoff another person can act on. The task sounds simple until the notes contain an ambiguous owner, an unconfirmed date, or a request the team never agreed to fulfill. Those are useful workshop materials because participants can practice judgment as well as prompting.

The agenda below is an illustrative 90-minute session for this task. It is not a report of a client workshop, a standard Jiberish Labs duration, or the scope of the existing sales Sprint. Adapt it to participants' experience, accessibility needs, approved tools, and available time.

Prepare two versions of the same task

Create a fictional or approved note set with clear actions and owners. Then create a second set containing an unresolved date, conflicting statements, and an action without an owner. Provide an example of a good handoff and the review criteria before participants begin.

Check access using an ordinary participant account. Confirm which information can be used in the approved tool. Supply the notes in a readable document, rather than making people transcribe a projected slide.

The exercise should end with an internal draft. It should not automatically message customers, update production records, or assign work during practice.

A 90-minute working agenda

If the opening access check consumes the practice time, shorten the scope or arrange another session. Racing through an exercise does not recover the learning opportunity.

Give participants instructions they can inspect

Use this example prompt with the supplied note set:

Create an internal handoff from the notes below.
Use only information present in the notes.

Return:
- A short summary of the agreed work.
- Actions with an owner and date where explicitly stated.
- Missing or conflicting information that needs clarification.

Keep proposals separate from commitments.
Do not assign an owner or date to fill a blank.
For each action, include the source phrase that supports it.

Notes:
[approved exercise notes]

Then inspect the source phrases. A requirement to cite the notes makes the draft easier to check, but it does not guarantee that the assistant followed the instruction.

For the ambiguous example, make “ready next week” appear near two different tasks. Ask participants whether the draft has attached it to the right one, whether it is a commitment, and whether an exact date is justified. This is a more revealing review than asking whether the prose sounds professional.

Evaluate the correction, not just the draft

A participant succeeds when they can explain what the output got wrong and produce an acceptable handoff. Use four criteria: supported facts, preserved uncertainty, clear ownership where known, and useful organization.

NIST describes confidently incorrect generated content as a risk in its generative AI profile. The exercise above is a practical way to teach checking behavior around that risk, not a certification exercise. NIST Generative AI Profile.

Collect the instructions, an approved example, and a short review checklist in one shared location. Name the person who updates them. Ask each participant to try the method on a suitable task afterward and record the corrections it required.

Scope the follow-up honestly

A workshop can reveal that the underlying notes are inconsistent or that staff lack access to necessary information. Those findings may call for process changes rather than more training. Optional follow-up can review actual outputs and decide which issue needs attention.

The published sales enablement example describes a sales-focused offering. Tailored AI Training for Teams can cover other departments and their tasks; its scope and estimate are agreed separately.

Use the workshop planning guide to prepare your task pack, or the team training guide to plan continued practice.