Multi-Agent Architecture: Your AI PM Team, For Product Managers

Priyanka Shetty

AI Product Leader & Builder

This course is popular

33 people enrolled last week.

Build Your AI Agentic PM Team

Every PM knows this Monday. A key metric drops, three people offer three different causes, and leadership wants an answer by noon.

We start the workshop with the decision you have to make. You work back from that decision to the outputs you need, those outputs tell you which agents to build and when you decide.

Then you build the team in Claude Code, in three hours:

  • Customer Intelligence finds what is happening & cites the evidence for every claim

  • Product Strategist turns that finding into options & one recommendation based on evidence

  • Product Challenger checks all the evidence & challenges the other agents

  • You, the PM make the decision

Finally, you connect the 3 agents. Give each one only the evidence it needs, check every handoff, stop the run when a check fails and end with the PM's decision.

You break the team on purpose: with a leading question, a request the evidence cannot support, and with an incomplete evidence pack. You score each agent and the team as a whole, write a five-line decision memo, and finish by designing your first agent.

The workshop runs on a realistic B2B SaaS case, FlowPilot, where new-user onboarding fell from 80% to 60%.

Understand. Recommend. Challenge. Decide.

What you’ll learn

Design, build, and validate a three-agent AI PM team you can adapt to your own product workflows.

  • Distinguish prompts, workflows, agents, and multi-agent systems.

  • Identify where judgment is required versus simple execution.

  • Decide where AI should act and where product judgment must stay human.

  • Define responsibility, context, instructions, guardrails, and output.

  • Give each agent only the context needed for its role.

  • Create outputs another agent can reliably use.

  • Build Customer Intelligence, Product Strategist, and Product Challenger.

  • Use realistic product evidence rather than polished demo data.

  • Keep each agent’s responsibility intentionally distinct.

  • Define what each agent passes to the next role.

  • Diagnose whether a failure comes from the instructions, context, workflow, or model.

  • Prevent context debt and role drift across the workflow.

  • Use break tests to expose unsupported claims and weak assumptions.

  • Separate supported facts, hypotheses, and unknowns.

  • Validate outputs before the next agent relies on them.

  • Use AI to understand, recommend, and challenge.

  • Make the final decision through a Human PM Decision Gate.

  • Apply the same architecture to your own product workflow.

Workshop agenda

  • Diagnose the PM Workflow and work back from the decision

    Decide whether the work needs a prompt, a fixed workflow, an agent or a team using a three-question test. Bring one decision you own, the outputs it needs and where it needs judgement.

  • Design and Build Your First AI PM Agent

    Write the agent yourself using Responsibility, Context, Instructions, Guardrails, and Output. Every claim it makes must cite the evidence and separate supported facts from hypotheses and unknowns.

  • Build the Strategist From Your Own Blueprint

    Complete the Agent Blueprint, then have Claude Code generate the agent from it. Compare your version with a reference build, and leave with a method you can reuse for any agent you design next.

  • Challenge and Break Your Team

    Add the Product Challenger and see why it needs different context and a different role than the Strategist. Run break tests, including a case where the right answer is "I don't know," and diagnose eac

  • Connect, evaluate and decide

    Run the full team, validate each handoff before the next agent relies on it, score the team on five checks. Make the call at the Human PM DecisionGate with a memo: Proceed, Revise, Investigate or Stop

  • Design Your Own AI PM Team

    Map the outputs your own decision needs, choose the roles, Blueprint your first agent, get peer review in small groups, and leave with a plan for Monday.

Learn directly from Priyanka

Priyanka Shetty

Priyanka Shetty

AI Product Leader (Oracle, S&P Global, Weill Cornell) | AI Strategist & Founder

Oracle
S&P Global
Weill Cornell Medicine
See all products from Pri

Who this workshop is for

  • Individual Contributor PMs building AI features
    Own AI initiatives but can't yet evaluate architecture, want hands-on build skills.

  • PMs targeting AI-native roles
    Job specs now expect agentic fluency. Want proof you've built one, not just read about it.

  • Product leaders scoping AI initiatives
    Need to evaluate engineering proposals with real technical judgment, not defer entirely or guess.

Prerequisites

  • No coding or AI engineering experience is required.

    You should be comfortable with core product-management concepts. Basic experience using ChatGPT or Claude would be good to have.

  • Set Up Your Tools

    Create your Claude Code account and confirm you can open it before the live workshop.

  • Test the Starter Project

    Open the provided project and complete the setup check to confirm everything works correctly.

What's included

Priyanka Shetty

Live sessions

Learn directly from Priyanka Shetty in a real-time, interactive format.

Lifetime access

Go back to course content and recordings whenever you need to.

1:1 with your instructor

Book your 1:1 when you enroll

Community of peers

Stay accountable and share insights with like-minded professionals.

Certificate of completion

Share your new skills with your employer or on LinkedIn.

Live Hands-On Workshop

Build one AI PM agent end to end, then connect it with two specialist agents using guided scaffolds.

Agentic AI PM Team Scaffold

Start with a reusable three-agent system designed around real product management workflows.

AI Agent Blueprint

Define each agent’s role, goal, inputs, reasoning, outputs, guardrails, evaluations, and handoffs.

Realistic PM Evidence Pack

Build and test using customer research, product data, and business context, not a simple demo.

Agent Templates and Prompts

Customize ready-to-use instructions for specialist agents instead of starting every build from scratch.

Testing and Evaluation Toolkit

Evaluate output, evidence, identify failure modes, and improve your agent team systematically.

Maven Guarantee

Your purchase is backed by the Maven Guarantee.

Frequently asked questions

Maven for Teams

Reimbursement

Get your company to pay

Everything L&D needs: email template, receipts, and certificate of completion.

Get reimbursed

Team discount

Learn with your teammates

Save 20%+ when 2 or more teammates enroll in the same cohort.

Save 20%+ with a team

Private cohort

Run a cohort for your org

A dedicated cohort with a custom schedule and curriculum, tailored to your team.

Book a private cohort

$350

USD

·
Dec 4
·

12–3pm EST

·

3 cohorts