Deepgram Raises $25M to Expand Enterprise Speech Recognition Platform
Speech recognition technology startup Deepgram has closed a $25 million Series B funding round led by Tiger Global with participation from Wing Venture Capital, Citi Ventures, SAP.io, and NVIDIA Inception GPU Ventures. The funding almost doubles the investment raised by Deepgram as it expands the reach and features offered by its enterprise-focused automatic speech transcription and analysis platform.
Deepgram’s technology gives enterprise systems voice transcription and analytics in real-time and in batches. The platform records and analyzes customer interactions, internal meetings and presentations, and any other conversation. Deepgram has long claimed better accuracy, greater than 95%, and higher efficiency for its automatic speech recognition using end-to-end deep learning and tunable models based on industry and environment compared to traditional heuristic models. The flexible nature of the platform and its ability to accommodate the quirks and vagaries of individual companies and industries combined with the accuracy attract enterprises looking to invest in voice services.
“Think of Twilio, think of Stripe, think of an API integration you use, which is a piece of a greater system you build yourself internally, rather than buying a behemoth one-size-fits-all instead,” Deepgram CEO Scott Stephenson told Voicebot in an interview about the new funding and Deepgram’s development. “It’s is a developer-first, go-to-market motion that we’re talking about here. If you have tens, hundreds of thousands, millions of users using a platform, you have to think about use cases. One-size-fits-all just won’t do it.”
If you have tens, hundreds of thousands, millions of users using a platform, you have to think about use cases. One-size-fits-all just won’t do it.
Deepgram processed more than 100 billion words in the last year, gaining more than 40 new enterprise clients, and hiring almost 50 people to keep up with its growth. The lengthening roster of clients is expanding the real-time transcription and analysis segment of Deepgram’s services, though the post-recording batch mode still makes up most of Deepgram’s work.
“We’re in the rise of real-time,” Stephenson said. “Many of the hurdles are getting moved out of the way and accuracy is improving. We had three times as many using the batch mode [over real-time], but that number a year ago was three times lower.”
The company closed a $12 million funding round last March and another strategic investment in June from the U.S. intelligence community-founded venture capital firm In-Q-Tel for an undisclosed sum. Deepgram began providing free, limited to its technology via the new MissionControl system and launched AutoML, a method used to get one artificial intelligence to train others without the need for humans to manually craft and adjust all of the models.
“Building an AI company, a deep AI company, is difficult compared to a standard SaaS company,” Stephenson said. “Difficult because you have to accrue your own data and label your own data. “We developed a way over the last year to train without using your own data. You just tell us what the keywords are. Behind the scenes, we generate data that is used to train the model.”
Praise from clients and Deepgram’s success over the last year likely helped open investor wallets for the new funding round. But there’s no question that the COVID-19 pandemic helped fuel the rise of enterprise voice services from every direction. A survey from speech recognition technology developer Speechmatics reported about two-thirds of enterprises now have a voice technology strategy jumping 18% from the year before. The pandemic has upped interest by the healthcare industry in Deepgram’s services as well, though Stephenson pointed to finance as a growth area the company will be focusing on in the future. Deepgram’s new lead investor is keen to get involved with the companies catering to filling the needs of these new enterprise voice strategies.
Dev building and data-driven decisions are going to define modern enterprise in the future
“Tiger is very excited about the voice space. One of the things that attracted them to Deepgram is they were looking a little deeper and understood the [value of the] tech powering voice apps,” Stephenson said. “We are the underlying infrastructure that powers it. How we work is we build our own data center. It’s how we can be so fast in development and how we can provide such a rock-solid service.”
The investment and acquisition of startups in the field show plenty of opportunity as of organizations incorporate speech recognition and transcription tech. Moments like Medallia spending $59 million to buy voice transcription startup Voci Technologies or business communications platform Dialpad raising $100 million to create voice AI services highlight how much potential lies untapped. Stephenson is confident in Deepgram’s position, however.
“Dev building and data-driven decisions are going to define modern enterprise in the future,” Stephenson said. “There hasn’t been a speech company that has been doing this before, making us [the winner] in how the west was won.