AI changed how we build. It hasn't changed why we build.
The fundamental product jobs — validating customer value, business viability, and usability — didn't disappear when coding agents arrived. They just need to keep up with the pace coding agents set. Andrew Ng has pointed out that PMs are becoming the bottleneck for AI-native teams. YC's 2026 Request for Startups put "Cursor for Product Managers" near the top of the list. Linear, Karpathy, and the spec-driven-development movement have all noticed agents need structured context, not narrative documents.
So we built the product side of that answer.
Today we're introducing the Revelica product agent: an AI teammate that handles customer research, identifies assumptions, generates ideas, and exports validated specs — at the speed your coding agent ships code.
How the product agent works
The product agent is built to run The Revelica Method — our opinionated framework for AI-native product discovery, organized into four phases: strategy, customer insights, solution validation, and delivery. The Method is the conceptual lens. The agent's actual work is more granular than that — and it's the granularity that lets discovery move at agent-speed.
The agent is powered by a library of skills and playbooks:
- Skills run one focused step. Analyze an interview transcript and extract opportunities. Ingest a research URL into the evidence base. Refine an idea against the desired outcome. Run a pre-mortem against a hypothesis. Apply Analysis of Competing Hypotheses to a decision. Rank a set of options. Generate user stories with acceptance criteria and risky assumptions from a feature description.
- Playbooks chain skills together for bigger jobs. Bootstrap a new workspace from just a company URL — research the company, surface customer segments, find and analyze competitors, articulate the value proposition, and produce a tested hypothesis. Or research a fresh customer segment end to end.
The library is the canonical list, and new skills land regularly as new patterns prove worth automating. The agent picks the right one for what you're asking, but you can also run skills and playbooks directly when you know what you want.
Either way, the agent works alongside you — you steer, decide, and bring the customer truth. It does the synthesis, citation, structure-keeping, and the parts that benefit from being run hundreds of times instead of once.
Discovery at agent-speed
Here's the failure mode the AI era exposed: shipping is fast, discovery is slow. So teams either skip discovery and let the market judge their assumptions, or they do discovery once and then stop — letting the original spec rot as new questions come up mid-build.
Both are how you end up with a customer-shaped hole in your product.
The product agent closes the gap. Discovery happens in the same flow as the build. You can run a fresh customer research pass on the latest interview transcripts at any point. You can generate three alternate solutions to the one your coding agent is currently building, see the assumptions each makes, and kill the one your evidence won't support. You can export an updated spec mid-sprint when the requirements actually shifted.
This is what we wrote about last month when we argued that agents need customer context, not narrative PRDs. The product agent is that context — built, maintained, and held to evidence by an agent that's on your team, not your customer's.
Why citation matters
Your coding agent will enthusiastically build whatever you describe. So will your customer research agent — if you let it.
The product agent's job is to refuse. Every claim about a customer needs evidence: an interview citation, a ticket reference, an observation. When the agent can't find the receipt for something it's about to assert, it surfaces the gap rather than papering over it. This is the discipline that prevents the operator-and-LLM bias loop that ships features nobody uses.
You and your LLM share the same blind spots. Structured evidence is what breaks you out of them.
What changes for Revelica
This is a relaunch as much as a feature. Revelica is now organized around the product agent. The skills and playbooks are the workflows the agent runs. Story maps, opportunity solution trees, experience maps, hypotheses, and assumptions are the structured artifacts it produces. The agent is the engine that connects them all to your team's real strategic context — your value propositions, customer segments, and current opportunities.
The output is structured discovery your team and your coding agents can both work from.
Try it now
Get started free. No credit card. Self-serve. Your first playbook run takes about two minutes.
If you'd rather kick the tires without an account, the Revelica story map free tool will show you what a story map looks like in five minutes, no signup required.
Feedback goes straight to our Slack. I can't help but drop what I'm doing to read them, even the hot takes.
Randall
