Working With Engineers, Legal & Leaders
How to partner with each group, write decisions down and plan under uncertainty.
A product manager succeeds by helping others work together. Share problems and evidence with engineers, bring legal and privacy in early, and give support, sales and leaders honest limits. Write one-page decision notes, plan around outcomes, and resolve conflict by naming the trade-off.
- Stakeholder
- Someone affected by the product or able to decide about it.
- Decision note
- A one-page record of a choice and its reasons.
- Outcome roadmap
- A plan built around results and evidence, not dates for features.
A product manager has no direct authority. You succeed by helping many different people work toward the same goal. AI adds new partners and new kinds of disagreement. Engineers disagree about what is possible. Legal worries about risk. Sales wants bold claims. This lesson shows how to work with each group, how to write decisions down and how to keep trust when outcomes are uncertain.
Working with engineers
- Bring the problem and the success measures, not only a solution.
- Ask for a rough plan early: what is easy, what is hard, what is unknown.
- Agree on the test set and pass mark together. Build the product against the evaluation.
- Accept that AI work is experimental. Plan time for trying things, and ask for evidence at set points.
- Respect technical limits. If the answer is it is not possible yet, ask what would make it possible.
Working with data and evaluation specialists
- Involve them in writing rubrics and building test sets.
- Ask them to explain noise, bias and what a score really means.
- Share real failures from users, so tests reflect real use.
Working with design and research
- Share the failure modes early. Designers can shape how doubt, errors and feedback look.
- Test prototypes with real users, including the confusing moments.
- Agree on tone and wording for the AI's voice.
Working with legal, privacy, security and governance
- Bring them in early, at the idea stage, not the week before launch.
- Give them the facts: data used, who is affected, what the AI does and what a person checks.
- Ask what conditions would make the idea acceptable. Treat them as partners in finding a safe way.
- Follow the tiers and review steps, and record decisions.
Working with support, sales and leaders
- Give support and sales honest limits and good examples.
- Warn leaders about uncertainty. Say what you know, what you expect and what you will learn by when.
- Show progress with real examples, including failures and fixes.
- Do not promise accuracy or dates that the evidence cannot support.
Decision documents
Write important decisions down. A short decision note saves arguments later and helps new people. Keep it to one page.
DECISION: {{what we decided}} DATE: {{...}} DECIDER: {{name}}
CONTEXT: {{the problem and what changed}}
OPTIONS CONSIDERED: {{A, B, C with pros and cons, in one line each}}
EVIDENCE: {{test results, user research, cost numbers; links}}
RISKS AND HOW WE WILL HANDLE THEM: {{...}}
WHAT WOULD MAKE US REVISIT: {{a number or an event}}
NEXT STEPS AND OWNERS: {{...}}Roadmaps under uncertainty
AI work is hard to date. Use outcome-based roadmaps: the problem to solve and the evidence you need. Use time boxes for experiments. State confidence levels. Update the roadmap as you learn. Leaders accept uncertainty when they see a clear learning plan.
Handling conflict
- Return to the user problem and the agreed measures.
- Separate facts from opinions. Ask for evidence.
- Name the trade-off aloud. Most conflicts are really trade-offs between speed and safety, or quality and cost.
- Escalate with options and a recommendation, not a complaint.
- ✓Work with engineers on problems and shared evaluation, not only solutions, and respect technical limits.
- ✓Bring legal, privacy, security and governance in early, with clear facts, and treat them as partners.
- ✓Give support, sales and leaders honest limits, examples and a learning plan.
- ✓Write one-page decision notes, plan roadmaps around outcomes and evidence, and resolve conflict by naming the trade-off.
Write a decision note and a stakeholder plan
- Pick a decision, such as launching an AI draft feature to all users or only to a beta group.
- Write the decision note with at least three options and a revisit trigger.
- List five stakeholders, what each cares about and what you will give them.
- Write the first two lines you would say to a worried legal partner.
Q1Why bring legal in at the idea stage?show answer
Early input shapes a safe design and avoids last-minute blocks. It also builds a working partnership.
Q2How do you plan a roadmap when AI results are uncertain?show answer
Plan around outcomes and evidence, use time boxes for experiments, state confidence and update as you learn.
Q3What makes a good decision note?show answer
One page with context, options, evidence, risks, a revisit trigger, next steps and owners.