Ethan Barhydt

The Dispatch

Most people hang up on a bad sales call. He recorded a few million of them.

Ethan Barhydt runs VoiceOps, a New York company built on a strange premise: the richest data your business owns is sitting inside conversations nobody bothers to study. Every day, call centers in insurance, real estate, travel and collections generate tens of thousands of phone calls. Most get scored once, by a tired manager, on a checklist. Then they vanish. Barhydt's bet was that those calls were a training set hiding in plain sight.

The framing he keeps coming back to is the self-driving car. Those teams feed their models millions of road images until the software knows a stop sign from a shadow. VoiceOps does the same thing - except the road is a conversation, and the stop sign is the moment a customer almost said yes. The product transcribes calls, structures the mess, and shows a coach exactly where what the rep should have said and what they actually said drifted apart.

It is, in the plainest terms, professional eavesdropping. The twist is who it serves. VoiceOps does not point the microphone at customers to sell them more. It points it at the salesperson, so they get better at their job before next Tuesday instead of finding out at the quarterly review.

Self-driving car companies take millions of images. We're doing the same thing - but instead of doing it on images, we're doing it on conversation. - Ethan Barhydt, on how VoiceOps thinks about AI

By The Numbers

What a listening machine does to a call floor

17→30%Conversion lift

150→8%Agent attrition

50,000Calls / day at scale

$9MSeries A, 2019

CONVERSION RATE · BEFORE17%

17%

CONVERSION RATE · WITH VOICEOPS30%

30%

Those are not vanity figures. An 80% jump in conversion on a floor that handles thousands of calls a day is the difference between a flat quarter and millions in new revenue. One client, the education group Penn Foster, posted the best conversion rate in more than a hundred years of operating. Early customers grew their VoiceOps spend from $50,000 to $250,000 a year - the surest sign a tool actually works.

Origin

A government major who shipped an AI model anyway

Barhydt studied government and politics at Harvard, which is a funny place to start for someone who would spend his career on machine learning. But he was already building. In college he wrote his first AI model and co-founded two startups - Jungol, which went through the 94labs incubator, and Valti, where he was COO. Between them he raised roughly $70,000 from angel investors and got the only education that matters for a founder: shipping things and watching what breaks.

The resume between dorm room and VoiceOps reads like a tour of how ambitious people figure out what they care about. Operations at General Assembly. A directorship at Harvard College Consulting Group. A fellowship at the Chicago Mayor's Office. A stint at the Center for American Progress. Public policy and startups, side by side, until the startup half won.

He founded VoiceOps in 2016 after seeing how badly call center teams coached their people. In early 2017 the company joined Y Combinator's Winter batch. By 2018 he was on the Forbes 30 Under 30 list for enterprise technology - the kind of stamp that gets a founder's calls returned.

The Arc

Career, in stops

COLLEGE
Builds his first AI model and co-founds two startups, Jungol and Valti, raising about $70K from angels.
PRE-VOICEOPS
Roles at General Assembly, Harvard College Consulting Group, the Chicago Mayor's Office and the Center for American Progress.
2016
Founds VoiceOps to bring AI to call center sales and collections coaching.
2017
VoiceOps joins Y Combinator's Winter 2017 batch.
2018
Named to Forbes 30 Under 30 in enterprise technology.
2019
Raises a $9M Series A led by Bain Capital Ventures; Accel and YC join. Ajay Agarwal takes a board seat.
2025
Raises a $5M seed, reframes VoiceOps around "AI coworkers," and talks shop on the Plugged In podcast.

The Idea Now

Coworkers, not replacements

The newer version of the pitch swaps the word "tool" for "coworker." Barhydt's argument is that good AI should sit next to a sales team and feed it real-time coaching, not quietly delete the team. Traditional coaching reviews one or two calls a quarter. VoiceOps reviews thousands, spots the pattern, and hands over feedback without the awkward human wince that makes people defensive about criticism. The machine has no ego, so neither does the note.

He is also blunt about how fast the ground moves. What the software can do shifts month to month, he says, so the only durable strategy is staying adaptable. The longer ambition runs past sales coaching entirely - toward conversation intelligence that reshapes how whole companies learn from the moments they currently throw away.

Unglamorous on purposeHe bet on the call center, the part of business everyone ignores - which is exactly why the data was untouched.

Small team, real revenueVoiceOps has reportedly done millions in revenue with a team you could fit around one table.

Politics to MLA government degree, not a CS one, yet he wrote AI models in college.

Feedback without the flinchThe product's quiet superpower is removing the human discomfort from giving criticism.

Watch

Hear it from him

Barhydt on the Plugged In podcast, on AI-driven coaching and the road from analyzing voice data to building AI coworkers:

How Ethan Barhydt and VoiceOps Are Transforming AI-Driven Coaching →