Inês Lourenço
CPTO and founder at Compound Works, Inês helps product leaders build AI-powered operating systems for their teams. She designs context layers, agent workflows, and decision frameworks that let PMs move faster, think clearer, and execute at a higher level.
36 guides by Inês Lourenço
AI skill design practice exercise for product managers
This exercise has you design one working slice of a personal AI operating system: choose a recurring PM task, build the...
Andre Albuquerque & Inês LourençoWhat Is a Personal AI Operating System?
A personal AI operating system is a structured layer built around a language model, context files, reusable skills...
Andre Albuquerque & Inês LourençoThe AI product requirements document guide
An AI product requirements document runs eight sections, and each has to prove something rather than describe it: that...
Andre Albuquerque & Inês LourençoHow to use context files for product specs
A context file gives an agent the standing background a new team member would get in week one: what the product does...
Andre Albuquerque & Inês LourençoHow to review an AI written PRD
Reviewing an AI written PRD means running four specific passes: provenance, edge cases, non goals and metrics. The...
Andre Albuquerque & Inês LourençoBest AI operating system resources for product managers
This is the reading list Builders Camp's Building your AI Operating System bootcamp curates on Claude Code, Claude...
Andre Albuquerque & Inês LourençoAI prompts for writing user stories
Seven prompts cover the real user story workflow: extract, draft, split, interrogate the reason, draft criteria, prune...
Andre Albuquerque & Inês LourençoHow to write release notes with AI
Feed the model merged pull request titles and linked issue titles rather than raw commits, group the output into the...
Andre Albuquerque & Inês LourençoHow to write a product one pager with AI
Amazon caps the press release half of a PR/FAQ at less than one page and its teams write ten drafts or more before...
Andre Albuquerque & Inês LourençoHow to write a PRD for AI features
An AI feature PRD adds four sections a normal one does not have: the eval set that defines good, the guardrails that...
Andre Albuquerque & Inês LourençoHow to write a PRD that gets built
Write a PRD in four passes: evidence, scope, requirements, then the cut line, and do not start the requirements pass...
Andre Albuquerque & Inês LourençoHow to write a product brief with AI
A product brief is one page and four sections: the problem, the evidence behind it, the direction you recommend, and...
Andre Albuquerque & Inês LourençoHow to write user stories with AI tools
Draft user stories with AI in three passes: give it raw evidence rather than your summary, make it propose a split...
Andre Albuquerque & Inês LourençoAI for executive summaries: writing for someone who reads 200 words
An executive summary is five sentences: what happened, the number, why now, what it cost, and what you need from the...
Andre Albuquerque & Inês LourençoAI for internal comms: the announcement is easy, the tone edit is the job
A model drafts an internal announcement in under a minute and gets the structure roughly right and the register wrong...
Andre Albuquerque & Inês LourençoAI for launch planning: build the plan from the scope, not from a template
A launch plan generated from a generic checklist is wrong before you read it, because it does not know what you cut...
Andre Albuquerque & Inês LourençoAI for meeting prep: walking into a decision meeting already argued with
Thirty minutes of prep changes a contested decision meeting more than an extra day of analysis: one sentence naming the...
Andre Albuquerque & Inês LourençoAI for negotiation prep: rehearsing positions before the room decides them
Negotiation prep with a model is two artefacts: a grid of interests and priced tradeables, and a rehearsal where the...
Andre Albuquerque & Inês LourençoHow to use AI for OKRs that survive contact with the quarter
The highest value use of AI on OKRs is the stress test, not the draft: ask for the data source and query behind every...
Andre Albuquerque & Inês LourençoAI for positioning and messaging: variants are cheap, the differentiator is not
A model will write eight positioning variants in the time it takes you to write one, and every one of them is available...
Andre Albuquerque & Inês LourençoHow to use AI for prioritization when an AI score is an argument, not an answer
A model will score 60 backlog items on RICE in about two minutes, and at least two of the four inputs it uses will be...
Andre Albuquerque & Inês LourençoAI for product marketing: five stages, and the two AI should not decide
Product marketing runs as a loop of five stages: market insight, narrative and positioning, product exploration, launch...
Andre Albuquerque & Inês LourençoAI for product presentations: structure first, slides second
A decision deck needs nine slides, and slide one is the ask. AI fills that structure from a source document in a single...
Andre Albuquerque & Inês LourençoHow to use AI for roadmap planning without outsourcing the sequence
The useful AI move in roadmap planning is synthesis, not sequencing: paste six categories of scattered input and get...
Andre Albuquerque & Inês LourençoAI for sales enablement: turning product truth into material reps will use
Sales enablement fails on usability, not volume: reps use material that is short, true and findable in the ninety...
Andre Albuquerque & Inês LourençoHow to use AI for stakeholder updates people actually read
Hand a model your notes and it drafts a complete weekly update in about a minute, which is the right division of labour...
Andre Albuquerque & Inês LourençoHow to manage context in Claude Code when you are a product manager
A Claude Code session degrades once the transcript fills with superseded decisions, and the fix is four habits: check...
Andre Albuquerque & Inês LourençoHow to triage customer feedback with Claude Code and keep every quote traceable
Triage works when every theme cites the row identifiers behind it, which means the export has to stay a file with...
Andre Albuquerque & Inês LourençoHow to build a launch checklist with Claude Code
Build the checklist from two inputs: the diff, which tells you what actually changed, and the spec, which tells you...
Andre Albuquerque & Inês LourençoHow to turn meeting notes into decisions with Claude Code
A decision record needs four fields per decision: what was decided, who owns it, what would reverse it, and the...
Andre Albuquerque & Inês LourençoClaude Code permissions and safety for product teams
Permission rules are evaluated deny, then ask, then allow, and a deny at any level cannot be overridden anywhere else...
Andre Albuquerque & Inês LourençoHow to verify AI-generated code as a product manager, before it merges or reaches a decision
Verify AI-generated code by asking for evidence instead of a summary: the test output, the build result, a screenshot...
Andre Albuquerque & Inês LourençoHow to synthesise a roadmap with Claude Code, and where the agent must not decide
An agent can assemble every roadmap input into one candidate list with sources attached and name the contradictions...
Andre Albuquerque & Inês LourençoHow to synthesise user research with Claude Code
Synthesise in two passes rather than one: code each transcript on its own, then cluster those codes across transcripts...
Andre Albuquerque & Inês LourençoMCP security for product teams, and the approval boundary to agree first
The MCP specification names user consent, data privacy and tool safety as key principles and then states plainly that...
Andre Albuquerque & Inês LourençoWhen not to use an AI coding agent
An AI coding agent earns its cost on work that is large, mechanical, well specified and reversible, which is a narrower...
Andre Albuquerque & Inês Lourenço