Synthesis Agent
Ranks problems by signal strength from evidence corpus.
What it does
Synthesis is the heart of the evidence loop. It reads all your evidence, clusters it into problems, scores each problem on real signals (frequency, severity, revenue impact, strategic fit), and ranks them — so the most important problems rise to the top, with every claim traceable to its source.
When to use it
Run it once you have a body of evidence and need to decide what to work on. It is the bridge from "we have a lot of feedback" to "here is the ranked list of what matters."
Turning a quarter of feedback into a ranked backlog
A PM inheriting a messy feedback pile
The situation. A PM joins a team with 400 open tickets, a churn survey, and sales notes — but no clear priorities. Everyone is arguing from anecdotes.
What they do
- 1They import it all and run Synthesis.
- 2It clusters the noise into eight problems and ranks them by composite score.
- 3The top three are grounded in severity and revenue, not just volume — with source evidence attached.
The result. In one run, the argument shifts from “I feel like X matters” to a ranked list everyone can see the evidence behind. The PM walks into planning with a defensible order.
What it can do
- Evidence synthesis
- Problem ranking
- Signal scoring
- Blind spot identification
What you give it
When you run this agent you can tune it with these settings (sensible defaults are provided):
What you get back
- ranked problem clusters, each tracing back to its source evidence
Every output lands on your canvas as a node you can edit, connect, and trace back to its source.
Examples
- Synthesize evidence into ranked problems
- Score problems by signal strength
- Identify evidence gaps
Lesson: garbage in, garbage ranked
- What it is
- Synthesis is only as good as the evidence you feed it. Thin or untagged evidence produces a shaky ranking.
- Why it matters
- It’s tempting to run synthesis on three tickets and trust the output. But the scoring needs enough signal to separate real patterns from noise.
Import a meaningful body of evidence (dozens of items, credibility set) before you lean on the ranking. Then re-run as new evidence arrives.
Related agents
Competitor Analysis Agent
Structured competitive analysis — competitor profiles, feature comparison matrices, market positioning maps, and strategic recommendations.
CSD Matrix Agent
Builds a CSD matrix — Certainties, Suppositions, and Doubts — to align the team on what it knows versus assumes.
Definition Agent
Synthesizes research into clear problem statements and requirements.
Empathy Map Agent
Clusters user research into distinct personas with goals, frustrations, and motivations.