Deepgram today revealed it has acquired OfOne, a provider of a voice platform that applies artificial intelligence (AI) to restaurants and other retailers that provide drive-thru services.
The platform will now provide the foundation for a “Deepgram for Restaurants” offering that leverages AI to improve customer experience and increase order accuracy, with additional integrations being added in the months ahead, says Deepgram CEO Scott Stephenson.
Fresh off raising an additional $130 million in funding, Deepgram has developed a set of frontier voice AI models that are exposed via a set of application programming interfaces (APIs) to enable speech understanding, speech generation, analytics, orchestration and autonomous voice agents.
The acquisition of OfOne now extends the reach of the company into a specific vertical market segment, which will soon be followed up by similar initiatives in other sectors, says Stephenson.
The overall goal is to make speech the primary interface for invoking a wide range of AI in much the same way Stripe is now used to process payments within a wide range of applications, he adds.
Adoption of voice AI technologies has been uneven, with noisy environments such as a drive-thru window being an especially challenging use case. “It’s an acoustically challenging environment,” says Stephenson. However, as the quality of voice AI continues to improve it has become feasible to, for example, rely on AI to process a customer order made at a drive-thru window, he adds.
Those types of use cases will become more common, especially once Deepgram is able to show its models can pass the Audio Turing Test at scale in 2026, notes Stephenson. The Audio Turing Test is designed to evaluate the degree to which an AI voice sounds human.
In theory, relying on voice AI should enable organizations to reallocate frontline workers to focus on other tasks, such as in the case of a drive-thru restaurant enabling employees to spend more time processing orders rather than taking them. Each organization will, however, need to determine how to reallocate their workforce versus simply reducing the number of employees they might hire to take an order.
Ultimately, it’s only a matter of time before voice AI becomes embedded in a wide range of applications involving everything from healthcare to manufacturing. Less clear is the degree to which speech might replace existing chat interfaces that still require end users to type prompts. The simpler the request, however, the more likely it becomes that a speech interface is likely to be preferred.
Of course, Deepgram is not the only provider of a platform that enables voice AI. Each of the major cloud service providers have their own initiatives. The challenge will be optimizing voice AI to the point where it becomes truly cost effective to deploy it ubiquitously in a way that doesn’t compromise a customer experience. Once that goal is achieved, however, it may no longer matter what language is being spoken by employees as voice requests are instantly translated.
Of course, there will still be customers that might prefer to chat with someone behind the counter while they, for example, place their order for a coffee but before too long those types of interactions may soon be the exception rather than the rule.
