Here is the business model half the sales coaching industry runs on. A rep gets on a call. The call gets recorded. On Thursday, a coach pulls up the transcript, scrubs to minute 14, and says: there — that's where you lost it. You should have asked about budget before you demoed.

The rep nods. The rep learns something. The deal is still dead.

That gap is the entire thesis behind Othello, a New York applied AI lab that announced $8 million in seed funding on October 14 from a16z speedrun, Owl Capital and Long Journey Ventures. The angel list reads like a founder group chat: DoorDash co-founder Andy Fang, Ripple co-founder Chris Larsen, Hugging Face co-founder Thomas Wolf, Sentry co-founder David Cramer, Writer co-founder Waseem AlShikh, Axonius co-founder Dean Sysman, Deel co-founders Alex and Philippe Bouaziz, and TransDigm CEO Kevin Stein.

Co-founder and CEO Jared Zelman's framing of the problem is the part that should make coaches sit up. He says there is a ton of technology before the conversation and far too much after it, but "not a single tool that affects how you perform during the call."

Read that again as a coach. Before the call: prospecting tools, CRMs, research. After the call: recording, scoring, transcripts, your Thursday review. Zelman is describing the shape of an industry — and announcing he intends to colonize the one part of it nobody has software for.

What it actually does

Othello listens in on video calls and uses AI trained on a company's uploaded sales and product materials to suggest questions to ask, answer tricky queries, and flag missed opportunities in real time through an on-screen overlay. Its proprietary model, Hamlet, is built to deliver live suggestions in under 500 milliseconds.

Half a second. That is roughly the length of a pause a buyer wouldn't notice.

It uses reinforcement learning to coach reps in pitches modeled on their company's top performers. Which is, functionally, a description of what a good sales coach does: watch the best person on the team, extract the pattern, install it in everyone else. The difference is delivery time.

There's a small detail from the Business Insider interview that works better than any demo. Zelman, now 24, was using Othello to help answer questions during the interview itself. He also offered the line the whole pitch rests on: "Too many good deals are killed by bad calls."

The numbers, and the asterisk

This is where a desk has to slow down.

The company says its customers include Microsoft, Capital One and Lyft. It says licensed reps adopt the app within a week of downloading it, and use it on 95% or more of their calls. Its a16z speedrun profile claims $600k in live ARR, $1.2M in cARR and $25M in enterprise pipeline since May, 30 enterprise customers all with market caps above $2B, and conversion rates improved by over 25% since launch. Elsewhere the deployment figure is reported as 50-plus multinational enterprises.

Those last two don't match, and the honest answer is that early-stage traction numbers move fast and get restated. Runtime Wire put the caveat plainly: "Its customer and performance figures remain company-reported," and the competitive field already includes vendors offering real-time coaching.

So: a real round, a real investor list, real named logos — and performance claims nobody outside the company has audited. Treat it accordingly.

What just got compressed

If your coaching product is send me your recordings and I'll tell you what went wrong, something in your offer stack just got squeezed. Not killed. Squeezed.

The squeezable parts are the mechanical ones. Did you ask a discovery question before pitching? Did you miss the buying signal at minute nine? Did you fumble the integration question because you didn't know the answer? That feedback has a half-life, and a machine that delivers it in 500 milliseconds beats a human who delivers it in four days. It isn't close.

There's a second-order effect worth watching. Gartner predicts that by 2028, 60% of B2B seller work will be executed through conversational user interfaces, up from less than 5% in 2023. If that holds even directionally, the enterprise buyer who used to hire a sales trainer for a two-day workshop will increasingly be buying a seat license instead. Budget moves from people to software. It usually does.

What it doesn't touch

Now the part Othello's own framing accidentally hands back to coaches.

The product gives reps the right words. It does not give them nerve. It can flag a missed opportunity; it cannot make a rep comfortable sitting in silence after quoting a price. It can surface the perfect objection handler; it cannot stop someone from discounting because they're scared of the pause.

Everything hard about selling is regulation under pressure. Ego, fear of rejection, the physical urge to fill dead air, the thing that makes a rep apologize for their own number. No overlay fixes that. A coach does.

There's also a quieter opportunity. Thousands of enterprise reps are about to get a live AI whispering in their ear for the first time, and most of them will be bad at using it — over-reading it, under-reading it, sounding like they're reciting. Someone has to teach people how to be coached by a machine mid-conversation without going robotic. That job did not exist on Monday.

The coaches who lose here are the ones selling transcript analysis as a premium service. The coaches who win are the ones who were never really selling information in the first place.