An AI strategy workshop for marketing agencies: a practical agenda

Sell a defined decision, not an AI performance

A workshop becomes difficult to evaluate when its promise is simply to unlock ideas. Define the decision before scheduling it: which segment deserves a first campaign, which claim needs stronger proof or which launch risk must be addressed before creative production begins.

An illustrative agency preparing a campaign for a B2B service might narrow the session to two positioning options. The client should know what will be decided, what will remain unresolved and which evidence is needed. Agree on the people who can approve the next experiment so the session does not end with attractive slides and no owner.

Prepare the evidence before the meeting

Ask for a concise product brief, the target buying situation, approved claims, known objections and relevant campaign or customer research summaries. Remove unnecessary personal or confidential information before using the material in an AI workflow. Mark sources and dates so disagreements can be traced back to evidence.

Prepare two scenario briefs with one important difference between them. Keep the offer, constraints and audience definitions consistent where possible. Run and inspect the Mirror scenarios ahead of the workshop; generation time and plan limits should not dictate the live agenda. Keep a written summary available if the demonstration cannot run.

• Decision owner: who approves the next action?
• Evidence owner: who can verify the claims in the brief?
• Two alternatives: what changes between the scenarios?
• Constraints: what cannot be changed during the campaign?
• Deliverable: what document will the client leave with?

Use a 75-minute agenda with clear outputs

The following agenda is a suggested facilitation format, not a measured promise about how quickly a strategy can be completed. Shorten the scope if the client arrives without enough evidence. A meeting cannot compensate for facts that nobody has collected.

• Minutes 0–10: restate the decision, success criteria and constraints.
• Minutes 10–25: review facts, assumptions and important unknowns.
• Minutes 25–45: examine contrasting scenario summaries and disagreements.
• Minutes 45–60: select the riskiest assumptions and feasible real-world tests.
• Minutes 60–75: assign owners, review dates and a condition for changing course.

Use disagreement as the working material

Show the scenario brief before showing the conclusions. Ask the client whether the represented roles and incentives reflect the market they know. A simulated reaction based on the wrong buying situation can sound plausible while being irrelevant to the decision.

When the client disagrees with an output, ask which evidence changes the interpretation. Record unsupported agent statements as discarded hypotheses. Do not count repeated simulated opinions as independent votes. The facilitator's job is to make assumptions discussable, not defend the tool.

Leave the room with a one-page decision memo

The memo should include the decision, supporting evidence, unresolved assumptions, chosen experiment and responsible person. Add a review date and a reversal condition: what observation would cause the team to stop or change the proposed direction? This makes the work useful after the presentation ends.

For the illustrative B2B campaign, the team might choose to test whether prospects understand an evidence-focused message before funding a larger media run. The memo should describe how comprehension will be assessed and who will review the results. It should not turn simulated enthusiasm into a sales forecast.

Prepare your next client brief in Mirror

Use the same structure for a first internal rehearsal: one client situation, two possible directions and the approved evidence behind them. Explore the scenarios in Mirror, inspect the report and extract the questions that require the client's judgment.

Make the source brief, scenario assumptions and action plan part of the handover. Clients should be able to distinguish their own evidence from AI-assisted interpretation. That transparency gives an agency a more professional deliverable and a practical basis for the next strategy conversation.

Common questions