Companies record everything now. Every sales call, every support chat, every email thread gets captured and transcribed, and then it just sits there. The data is right in front of you and almost none of it is usable. If I wanted to answer something as basic as "did the buyer confirm budget on this call" across two hundred deals, the only real answer was to get a human to read the transcripts. That isn't conversation intelligence, it's conversation storage with a nicer front end.
The tools on the market didn't fix this. They gave you a prettier summary and another dashboard, when what I wanted was the underlying data. So I built it. Semarize is an API.
What it does
Semarize is a conversational intelligence API. You send it a transcript - a call, an email, a chat, whatever - and it sends back structured, typed data. Not a paragraph for someone to read, but fields your systems can use directly. A boolean for whether a next step was agreed. A score for discovery quality. An extracted string for which competitor came up. Every field comes back with a reason and the exact span of the conversation it was pulled from, so you can audit it instead of trusting it blindly.
The output is just JSON:
{
"discovery_quality": 68,
"budget_confirmed": false,
"next_step_confirmed": {
"value": true,
"reason": "Buyer agreed to a technical demo on Friday",
"evidence": ["...let's get the whole team on a call Friday."]
},
"competitor_mentioned": "Gong"
}
That lands in a CRM field, a BI table, a Slack alert, or an automation step, the same shape every time. No parsing, no prompt babysitting, no one reading transcripts.
How you actually use it
Two concepts, that's it.
A Brick is a single semantic check that returns a predictable typed output. "Was a quantified pain identified." "What objection was raised." "Did the rep quote current pricing." You define what matters to you; the NLP side is handled.
A Kit is a reusable, versioned bundle of Bricks for a stage, team, or workflow. A discovery Kit, a MEDDICC Kit, a compliance Kit. You point a Kit at a conversation through the API and get every field back in one run.
No ML engineers needed. You write the logic in plain terms, version it like you would anything else, and call it on every conversation that comes through.
The bit I care about most
You can ground the whole thing in your own knowledge. Upload your playbooks, pricing, and product docs, and Bricks evaluate against your business rather than generic assumptions. A Brick can tell the difference between a rep quoting the live rate card and a deprecated one. Keyword matching never could, and a summary model just makes something up that sounds plausible. When your pricing or messaging changes, you refresh the docs and the evaluations move with you.
Who it's for, and who it isn't
It's for people who want conversation data as infrastructure. RevOps wiring CRM enrichment, GTM engineers building internal tools, data teams who want clean signals in the warehouse. It plugs in behind whatever you already record with - Gong, Zoom, Teams, Fireflies, your own pipeline - so there's no new bot and no rip-and-replace. The same engine runs across enterprise sales, financial services compliance, and contact centres.
It is not a meeting recorder, a coaching dashboard, or a CRM. If what you want is a rep-facing UI to rewatch calls in, Semarize isn't that, and I'd rather tell you now than waste your trial.
Try it
If any of this matches a problem you've got, the fastest way to see it is to throw a transcript at it and look at what comes back. Start for free, no card, testing in a few minutes.
Happy to answer anything in the comments.
