It's Tuesday. Somewhere, a Zoom call ends.
A sales rep in Austin closes the laptop, walks to the kitchen for water, and by the time she gets back, three things have already happened. There is a summary of the call in her inbox. There is a draft follow-up email waiting for her signature. And Salesforce - the application that has spent fifteen years training salespeople to resent it - has been quietly updated. She did not type a word.
The thing that did the typing is Sybill. It is not a notetaker. Sybill's founders are very particular about this. A notetaker writes things down. Sybill writes things down, decides which of them matter, files the ones that do in the right field of the right CRM record, drafts an email in the rep's own voice, and then - because someone has to - flags the deal as quietly slipping.
The CRM was supposed to help. It mostly didn't.
For about two decades, the dominant productivity story in B2B sales has been this: buy a CRM, hire a RevOps team, and ask reps to please, for the love of pipeline hygiene, update the fields. Reps, who got into the job because they like talking to people, do not love updating the fields. So the data is half-true, the forecast is half-baked, and every Monday morning a manager has to ask three follow-up questions to figure out what actually happened on a deal.
Sybill's founders watched this loop closely. Before they built anything, they interviewed roughly six hundred sales reps. The reps did not ask for more analytics. They asked for someone to do the paperwork. Sybill took the request literally.
Four roommates, one oracle.
Sybill began in 2020 inside a Stanford dorm-adjacent apartment, where four AI researchers - Gorish Aggarwal, Nishit Asnani, Soumyarka Mondal, and Mehak Aggarwal (Gorish's sister, formerly of Harvard and MIT) - were trying to build software that could read non-verbal cues in remote classrooms. The premise was unobjectionable. The pandemic was forcing teachers to lecture into rectangles of dark squares; surely a machine that could tell when students were lost would help.
The teachers were polite. They were not buying. Salespeople, on the other hand, were doing exactly the same job - reading rooms over video - and they had budgets. The team pivoted, kept the computer-vision research, and pointed it at the highest-stakes Zoom call in any company: the one with a customer on the other end.
The name, for the etymologically curious, is borrowed from the Sibyls of ancient Greece - oracles who whispered hidden truths. The branding is more restrained than that suggests.
What Sybill actually does, in plain English.
The product divides cleanly into two halves. The first half listens. Sybill joins a call - on Zoom, Google Meet, or Microsoft Teams - records and transcribes it, watches the body language on the customer side, and produces a Magic Summary the moment the call ends. The summary is structured: what was discussed, what was decided, what is supposed to happen next, and which buyer in the room sounded most engaged.
The second half acts. It opens Salesforce or HubSpot and updates fields that would otherwise have taken the rep twenty minutes to fill out. It drafts a follow-up email in the rep's voice, with the right customer-specific detail. It pushes risk signals into Slack when a deal stalls. For managers, it pulls back to the pipeline level and inspects deals the way a senior rep would, flagging the ones that are wobbling.
Magic Summary
Action items, next steps, and a paragraph of context, delivered before the rep's coffee cools.
CRM Autofill
Salesforce and HubSpot fields populated automatically from what was actually said.
AI Follow-up
Drafted in the rep's tone, with the customer's exact concerns referenced.
Deal Inspection
Risk detection and buyer intent across the whole pipeline, not just the call.
Sales Coaching
Behavioral analytics managers can actually act on, instead of file away.
A Compressed History
The numbers, mostly the company's own.
Roughly thirteen and a half thousand sales reps now use the product. They have, between them, fed it about twenty million minutes of customer conversation - which is, by any reasonable estimate, more than a hundred-year working life. Sybill calculates it has handed back about five million minutes of admin in return. That is the kind of statistic that sounds suspicious until you remember how much time a quota-carrying rep spends inside Salesforce text fields.
The 2023 growth curve is the part investors point to. Sybill scaled from one hundred thousand dollars in ARR to one million in nine months. Greycroft led the Series A on the back of it; Neotribe, Powerhouse, and Uncorrelated Ventures all came back from the seed.
Sybill, in four numbers
To sell is human. To solve is Sybill.
The internal slogan is, by company standards, almost old-fashioned. It is also a deliberate inversion. Daniel Pink's To Sell Is Human argued that selling is a fundamental human act. Sybill agrees - and then quietly suggests that the rest of the rep's job, the part with the dropdowns and the email templates and the Monday forecast meeting, is not a fundamental human act and could probably be done by something else.
What you do with the time Sybill gives back is, of course, your problem. The company would prefer you spend it on a customer. Or, failing that, lunch.
Why it matters past this quarter.
The conversation intelligence category is a decade old. The first wave - Gong, Chorus, and the rest - shipped dashboards. They told managers what reps were doing. They did not, in any practical sense, do the work for the reps. Sybill belongs to a second wave, one in which the AI is not a mirror but a coworker. It does not just describe the deal. It updates it.
If that wave plays out the way its early customers seem to believe, the CRM will stop being a thing reps grumble about and start being a thing reps barely notice. The next generation of sales software will probably look less like a database with a UI and more like a colleague with a Slack handle.
Back to Tuesday. The rep in Austin opens her inbox. The summary is there. The follow-up draft is ready. Salesforce shows the deal at Stage 3, with the right next step in the right field. She edits two sentences in the email, hits send, and joins her next call. The whole admin lifecycle of the previous meeting has taken her ninety seconds. Sybill is already in the new one, listening.