There’s a poorly kept secret about the effectiveness of voice AI systems in the drive-thru. Secondary noises coming from the car are a real pain.

Umut Isik, CEO of Incept AI
“In previous versions of the technology, the AI agent would get confused by interfering speech,” said Umut Isik, CEO of Incept AI, in an interview. “Kids are screaming, the radio is on loud. Or sometimes the agent will hear itself. You know how we hear ourselves in Zoom calls sometimes? The drive-in is basically a ginormous speakerphone. The agent will hear itself and confuse itself. All this can affect reliability.”
Isik, who has a background in applied science and previously worked at Amazon Web Services, is confident his new and improved voice bot can increase the clarity of those drive-thru orders, as well as phone and kiosk orders. And he has data to back it up.
“Our system completes 97 percent of orders accurately, without human help,” he said, noting that competitors often come in around 83 percent and require flesh-and-blood back-up.
The company launched in January 2024 and already boasts a couple 2,000-unit brands as pilot clients. One uses it in the drive-thru and the other for phone orders. “We can integrate with a client’s infrastructure providers, its point of sale system, its menu boards,” he said. “It’s really not hard to set up.”
Isik’s background is in building neural networks for audio processing and he’s one of many PhDs at the company.
The company has captured the attention of investors. Last month it announced it had raised $3 million in pre-seed funding from Rally Ventures.
Isik is good at an elevator pitch and frames his company’s unique selling proposition this way. “We solve for the last-mile problem in voice AI,” he said.
The company does this by building on foundation models, which are the building blocks in platforms like chatGPT.
“Foundation models are really quite amazing. We’re not building everything from scratch. We complement them with our own neural networks. We pre-process the audio to make it easier for the foundation model to work with the audio at the drive-thru. We allow the AI to listen better,” he said.
But that’s just on the audio side. There is still the question of how to handle consumer customization.
“Brands are quite proud of their customizability,” he said. “The question for a tech vendor is do you actually support all the customizations without issues? That’s a big piece of the puzzle. This is something we can do.”
Isik is betting that brands will find the improved accuracy of orders worth the tech upgrade.
“Customers expect and deserve reliability in their orders,” he said. “The audio processing is the key thing that’s going to make this happen.”