Skip to content
This site demonstrates one possible use of this domain. For acquisition, partnership, or investment inquiries, please use our contact form.
Workshopping.ai

Flagship workshop

AI Planning Framework Workshop

A facilitated working session that turns a goal into an AI-assisted plan you can actually execute — with humans owning judgment and AI accelerating the rest.

Last reviewed:

Most AI planning fails the same way: a single prompt produces a single plausible plan, the team nods, and ninety days later the plan has quietly become a wishlist. This workshop fixes that by forcing divergence, adversarial review, and a written human decision before anyone commits a calendar. AI does the heavy lifting on generation and critique; humans own the outcome and the accountability.

  1. 1

    Define the outcome, not the output

    Write the goal as a measurable change in the world, not a list of deliverables. 'Launch a course' is an output; 'Have 50 paying students by Q3' is an outcome. AI is excellent at generating outputs and terrible at choosing outcomes — humans must do this step.

  2. 2

    Inventory inputs the AI can use

    Gather every artifact the model should know about: customer interviews, prior plans, brand voice, constraints, budgets, deadlines. Put them in a single working doc. Models reason better with concrete context than with vague prompts.

  3. 3

    Generate three divergent plans

    Prompt the model to produce three genuinely different plans — conservative, aggressive, and contrarian. Ask it to name the assumptions each plan depends on. Divergence beats a single 'best' plan because it surfaces hidden trade-offs.

  4. 4

    Stress-test with a red-team prompt

    Feed each plan back to the model with: 'You are a skeptical operator who has seen this fail. What kills it in 90 days?' Capture the failure modes. This is where AI earns its keep — it has read more post-mortems than any human.

  5. 5

    Human decision, written rationale

    Pick one plan, or synthesize across them. Write a one-page rationale: what you chose, what you rejected, and why. The model can draft this, but a human signs it. Accountability does not delegate.

  6. 6

    Convert to a 30/60/90 with owners

    Translate the plan into milestones with named owners and weekly metrics. Schedule a 30-day review where you re-run steps 3-5 against fresh evidence. Plans are perishable; the framework is durable.

Worked example — a solo founder launching a B2B SaaS

  1. 1.Outcome: '20 paying teams at $99/mo within 6 months.' Not 'build a product.'
  2. 2.Inputs: 14 customer interview notes, a competitive teardown, a $12k budget, a 20-hour-per-week constraint.
  3. 3.Three plans generated: (A) Niche down to legal ops, hand-build for 5 design partners. (B) Self-serve PLG with a free tier and content engine. (C) Productized service first, software second.
  4. 4.Red-team: Plan B dies on CAC; plan C dies on founder burnout; plan A dies if the legal-ops niche is too small. Founder commissions a 5-call validation sprint on niche size before committing.
  5. 5.Decision: Plan A, conditional on validation. Rationale doc names the kill criteria: fewer than 3 of 5 design partners sign LOIs in 14 days = pivot to plan C.
  6. 6.30/60/90: Day 30 — 3 LOIs signed, prototype shipped. Day 60 — 5 paying teams. Day 90 — repeatable sales motion or pivot.

AI Planning Framework Workshop - FAQ

How is this different from just asking ChatGPT to make a plan?
A single prompt produces a single plausible plan. This framework forces divergence, adversarial review, and a written human decision. The difference shows up at day 60, when the single-prompt plan has quietly drifted and this one has named kill criteria.
Which AI model should I use?
Any frontier reasoning model works (Claude, GPT, Gemini). The framework matters more than the model. If you have access to multiple, run the divergence step on one and the red-team step on another to reduce shared blind spots.
Can a team run this together?
Yes — and it gets better. Assign one facilitator to drive prompts, one scribe to capture rationale, and rotate the red-team role. Two hours for steps 1-5, one week to execute step 6.
What if the AI's plan is obviously wrong?
Good. That is signal, not failure. A wrong plan is faster to correct than a vague one. Write down why it is wrong; that note is now an input for the next iteration.
How often should we re-run the framework?
Every 30 days for active initiatives, or whenever a kill criterion fires. The framework is cheap to re-run because the inputs doc compounds — each cycle leaves better context for the next.

Sources & further reading

  1. Anthropic — Building Effective Agents — Anthropic
  2. OpenAI — A Practical Guide to Building Agents — OpenAI
  3. Thinking, Fast and Slow — Daniel Kahneman
  4. Superforecasting: The Art and Science of Prediction — Philip Tetlock & Dan Gardner

Related

Framework

Human + AI Planning Framework

Workshop

How to Workshop an Idea

Workshop

Business Plan Workshop

Tool

Goal Planner