Brief (a16z sr005)

Product context is like an onion 🧅 At the top is a decision. “We’re using Stripe for payments” But that’s not really much context. Behind that is a why. “We need to ship fast, can’t build custom billing” But keep peeling back the layers and there is even more: “We’re launching in 2 weeks, and there are 15 P0 issues to fix first” And more: “When we talked to customers, the number one requested feature around billing is annual billing plans, which Stripe supports” To truly let agents build the right thing, you need all of this context. Which is why we built Brief. It navigates your agents to all of this context. So when you’re building in your repo, it knows not to suggest ripping out Stripe and all the layers of why behind that decision. Brief stops agents from building blind.

Context is only useful if it’s accessible to everyone on your team, including your AI agents. Markdown files are a good start, but that’s only half the picture. Which is why Brief is built as an end-to-end workflow that’s connected to your agents via CLI. Slack, Notion, Granola, Google Drive, Gong, and more are all accessible to your agents through the Brief CLI. Midway through working on a new feature and not sure which direction to go? Claude Code can use the Brief CLI to ask Brief what customers said on user research calls to clarify what they needed from this new feature. It can look through your documentation for all the tickets related to this task and auto-tag them in the PR. Brief navigates your agents to every bit of context they need so they stop building blind.

Most teams that use AI keep detailed documentation. Whether it’s git-committed .md files, or connecting AI tools to Linear and Notion. Part of making agents productive is giving them enough context to do their job properly. But this creates problems. For one, stuff gets out of date quickly. AI can certainly help in the culling and editing of stale docs, but it becomes a full-time job, almost a librarian role, to keep everything up to date. Additionally, a final direction is never the whole story. It’s one thing to know that your app uses Clerk for auth… it’s an entirely different thing to know the company history of auth providers and all the failed integrations along the way. This is our secret sauce around Brief. We keep a living web of all the decisions and documents you’ve made and create connections. That way when an AI agent goes to correct an auth bug, it doesn’t suggest switching to an auth provider to fix the problem that you dumped 6 months ago. This problem compounds as a company gets older. As people come and go from a company, context gets lost; you don’t remember WHY someone decided to do something. We built Brief to be a product that a new employee can ask about the entire company history around certain decisions and get the whole story on day one.

There are plenty of AI tools that help make tickets. There are plenty of tools to make PRDs. But that’s not really what a PM’s job is. A PM is a glue role, connecting stakeholders across the CXO team, Product, and Marketing, and figuring out how to reconcile their conflicting goals. Imagine a sales team wanting to add granular user settings to help close a deal. They’re talking to a customer who has particularly nasty RBAC requirements. Then you have the design team advocating for less granularity, arguing that for every feature added to user settings, things get harder to use. You start having roles like “super extra admin” that require a whole page to explain what they do and do not have access to. Not to mention a wall of checkboxes to give and take away permissions from different user types. This, in essence, is the real job of a good PM: to be the glue that navigates a feature through this minefield of conflicting priorities. This entire workflow is missing from today’s AI tools. They can certainly write tickets and PRDs, but they have no way to handle conflicts like this. Which is why we built Brief. Not to replace PMs, but to enable them to move at the speed of other AI tools. Brief already has all the sales calls with that customer who has particular RBAC requirements. It knows what your design team’s priorities are and the Slack messages around their concerns. It allows PMs to have all this context at their fingertips when trying to get a feature to release. Brief navigates an idea from initial exploration to production. Align your team and agents from vision to impact at AI speed.

One person, a thousand agents. Every AI startup is pointed in the same direction: make teams smaller, and make every employee 10x as effective by giving them a swarm of agents to work with. There’s no way this is going to work with today’s tech. Current AI tools are operating in a bubble. They know some of what they’re building. The repo gives some context hints. But imagine how much context would be missed if you hired a new employee, gave them the repo, and only provided additional context when they asked for it. Now apply this to one employee with a swarm of agents. What started as a way to make them work 10x faster turns into a constant stream of incoming questions and context that the bots need to do their job. Or the bots just get it wrong, usually badly. Not sure which is worse. This is why we built Brief. Agents need to stop building blind. Brief navigates them to the context they need (your customer calls, your ICP, your Slack threads) all programmatically through a CLI so they can execute on their own. We even ran a blind test of this 1:1 against Claude Code in a sandbox. 8/8 features completed with Brief vs. 2/8 without. Read the full paper in the link below 👇

See us at Web Summit Vancouver 🇨🇦 May 12–13, come say hi! 👇 Check the comments for a unique referral link (20% off General Attendee tickets)

Align your agents Product decisions in every IDE. Every builder aligned. PM in SF makes a call at 3pm. Engineer in Berlin picks it up at 9am. Every decision is instantly available across timezones, tools, and sessions. No waiting. No asking twice. Running 3 Cursor sessions in parallel? Brief navigates them all to the same product truth. No more copy-pasting prompt chains into Notion to replay in new sessions. A code change is about to violate a product decision? Brief flags it before it merges. Business context as a CI check.

AI coding tools transformed engineering. Brief does the same for product. Brief syncs from Slack, Linear, Notion, and GitHub automatically. Your context is always fresh. No more maintaining the Rube Goldberg machine of Fathom to Zapier to Google Docs to Notion you built last month. Ask any product question. Get answers with charts, cited decisions, and real data. Not a search engine. Answers that cite their sources. Someone needs your input at 2am? Brief navigates them to the answer with your voice, your context, your reasoning. Write tickets that actually get built right. Brief navigates your AI coding agents to the context they need to build what you meant.

Ship at the speed you think. Brief lets you capture context from anywhere. Slack message, whiteboard photo, voice note. Brief turns chaos into a PRD. Then generate wireframes and fill out details with info directly from your codebase. You edit instead of starting from scratch. Then it's just one click to go from PRD to Linear tickets with detailed acceptance criteria. And in the meantime, every call you make gets captured with rationale. Six months from now, you'll know exactly why you chose this direction over another. Brief navigates the entire process from idea to code so nothing gets lost along the way.

Al ships a million features. Pointed in a million directions. We open-sourced a benchmark that measures how often frontier models honor your team's decisions. The scores aren't flattering. This is the gap Brief closes. Link in comments, try it out with your own prompts.