The GTM brain and semantic graph — how Yuzu reads a deal | Yuzu
The GTM brain
Reading the call is easy. Reading the deal is not.
The GTM brain holds every call, deal, doc and price you have. The semantic graph on top knows what any of it means for the deal in front of you.
| A recorder | Claude + a connector | Yuzu | |
|---|---|---|---|
| Knows what was said on the call | ✓ Yes | ✓ Yes | ✓ Yes |
| Knows which of 400 deals matters today | — | — | ✓ Yes |
| Knows why now, and can show its work | — | — | ✓ Yes |
| Knows what closed your last ten deals | — | — | ✓ Yes |
| Writes the asset that fits the moment | — | Roughly | ✓ Yes |
| Stays quiet when nothing changed | — | — | ✓ Yes |
| Every agent reads the same current truth | — | — | ✓ Yes |
The difference is the ontology. The rest of this page is how it works.
Master brain
live
38
Items in the brain
34 current
7
Agents connected over MCP
4 tools
82%
Agent reads on current truth
| Agent | Job | Reads | Status |
|---|---|---|---|
| Claude | call prep | 612 / wk | current |
| Gamma | deck gen | 418 / wk | current |
| Codex | scoping | 280 / wk | current |
| SDR agent | outreach | 505 / wk | current |
| Pricing bot | quotes | 94 / wk | stale |
Use the seats you already pay for
No second AI bill. Your existing Claude and Codex subscriptions read the brain over MCP and get your full deal context.
Any rep, any tool, same answer
A rep asking Claude and a rev-ops lead asking Codex get the same deals, the same history, the same reasoning.
You can see what is stale
Which agents read what, how often, and which one is quietly quoting last quarter’s pricing.
The brain itself. What Yuzu knows about your go-to-market, with who last touched each item.
Govern your agents for revenue.
Concepts
The things you sell on.
Stages
How your deals really progress.
Signals
What moves a deal, learned.
Buying patterns
Who decides, and how.
Every closed deal sharpens the read. The curve is why month six beats month one.
Two deals, the same summary on day 0. The trajectory is what tells them apart.
None of this was configured. It is learned from the deals you already closed.
Champion is going quiet
78% confidence
Traces back to
- 0.31Champion email unanswered, six days
- 0.27Skipped the rescheduled call
- 0.22New legal stakeholder, no reply
A champion going quiet is a signal. A moved meeting is noise. Yuzu learns which is which from your own deals, then shows its work.
A read on the deal, with the receipts.
Every tool asks constantly. Yuzu acts once.
Detecting risk is easy. The hard part is that a seller can only act a few times a week. So Yuzu stays quiet on the deals that can wait, and spends the move on the one where it changes the outcome.
Call recorder
GongFathomGranolaZoomor ours
CRM
SalesforceHubSpotAttioor ours
Warehouse and docs
SnowflakeDatabricksDriveNotion
Agents, over MCP
ClaudeCodexGammaSlackyour own
Claude is the engine. Yuzu is built for GTM, so it works on day one instead of after a three-month science project.
What is a GTM brain?– A GTM brain is a single store of everything a company knows about how it sells, with a reasoning model on top. It holds calls, deals, docs, pricing and positioning, and understands what they mean for a specific deal rather than just retrieving them.
How is this different from connecting Claude to our Google Drive over MCP?+
Is this a vector database or RAG?+
Do we have to connect every tool separately?+
Which agents and tools can read it?+
Do we have to move off our CRM?+
Who owns the data in the brain?+
How long until the brain is useful?+
See it on your deals.
Connect your calls and CRM. We'll show you the patterns and the next move.