Startup Field Guide by Unusual Ventures: The Product Market Fit Podcast · Unusual Ventures

Deepgram's CEO Scott Stephenson on Speech AI

March 24, 2024·41 min·3 clips
Google and Microsoft laughed at Scott's deep learning pitch, so he started a $250M voice AI company.
The episode opens on the next frontier of speech AI: Scott Stephenson argues that the hard part is not just making bots speak, but making them respond in a way that feels coherent, empathetic, and human when the conversation turns negative or messy. Sandhya then brings the conversation back to Deepgram’s origin story, asking how a particle physicist ended up building the company and what practical problem the team chose to solve. The early market was narrow. At the time, call centers were the main buyers of speech recognition at scale, using it for analytics, script compliance, and figuring out whether customers were happy or frustrated. The show stays close to the everyday work of managers who used to sample only a few calls per person each week, and Deepgram’s pitch was to automate that feedback loop. The go-to-market shift matters just as much: Scott describes moving from free beta usage to paid usage and then to a real sales motion as the point when the business started to look repeatable, which led them to hire Chris Dyer, now Deepgram’s VP of sales, and raise an A. Sandhya keeps the interview anchored in early customers and memorable feedback, and the closing stretch adds a quick update that Nova 2 is out now and Aura is Deepgram’s low-latency, highly expressive text-to-speech model coming out. The result is a founder interview with a clear spine, more interested in the sequence of decisions than in mythology.

As heard by us

Deepgram's story here is about turning speech recognition into a repeatable go-to-market motion.

This conversation traces Deepgram from Scott Stephenson's particle physics background into voice AI and the harder problem of finding a repeatable market.

Read the full review in PlayNext →

Why you'd press play

How Deepgram turned call-center demand into a repeatable go-to-market motion after proving it with beta users.

Read the full recommendation in PlayNext →
Listen to the show on