The Story of Canary Speech: The Company Building the Future of Voice-Based Disease Detection

The Story of Canary Speech: The Company Building the Future of Voice-Based Disease Detection

The best companies begin with a single question that refuses to fade. For Canary Speech, that question emerged two hours into a conversation in a bagel shop in Provo, Utah.

In a recent episode of Category Visionaries, Henry O’Connell, CEO and Founder of Canary Speech, shared how one question to his friend Jeff Adams sparked an eight-year journey to build technology that detects diseases from voice—no words required.

The Question That Started Everything

Henry and Jeff had known each other forty years. Jeff had just finished leading the Amazon Echo team and had previously created Dragon NaturallySpeaking at Nuance. They sat in that bagel shop for six hours, discussing what to do next.

Two hours in, Henry asked: “Jeff, you know, recognizing the impactful career he had, I said, what do you want to do at this point? How would you apply what you have done and the skill and the expertise you have?”

Jeff’s answer: “I’ve always wanted to use speech and voice to identify human condition and disease and make an impact.”

The Insight That Changed Everything

Henry shared a personal observation with Jeff—one every parent recognizes. His daughter Caitlin would come home from high school soccer. “Just from her gait, as I mentioned earlier, you know, multimodal biomarker analysis that our brain is doing from her gait, I knew it was a good day or a bad day.”

Years later, when Caitlin was married with a child, Henry would call her twice weekly. “When she answers the phone, and it has absolutely nothing to do with the words she’s speaking, and it’s irritatingly accurate. And I don’t have the physical cues anymore. I can’t see her walking. I can’t see her face, but I know whether it’s a good day or a bad day.”

He asked Jeff how we do this. Jeff’s response changed everything: “I don’t know, but it has little or nothing to do with the words that people are actually speaking.”

The Problem Everyone Was Solving Wrong

For thirty-five years, researchers tried using speech to detect disease by analyzing words. The entire field was thinking within a box defined by natural language processing. Jeff and Henry looked at what everyone ignored: sub-language elements. The pitch. The rhythm. The trembles.

“Jeff likes to call it the musical component or elements of speech,” Henry explains. “You know that I’m excited about what I do, and I’m not telling you I’m really excited about what we do.”

Building What Universities Couldn’t

The first years were about survival. They took project work with UnitedHealthcare. “We actually did multiple ones with United Healthcare, probably eight of them, which really helped our company.” By year two, they hit a million dollars.

But customers would approach them, then compare them to universities. “At some point, we would be approached by a pharmaceutical, or we’d be approached by a healthcare company to do that kind of work. They had friends that were, or they were graduates of a university, and they would go to their university.”

If MIT could do what Canary Speech was doing, they weren’t building defensible technology. The solution: intellectual diversity. “In the past, we had two very bright individuals, but they were from the same graduate program.” They hired from the UK, Germany, Korea, and the US. “You couldn’t think within the box, but because for the last 35 years, thinking within that box had not produced a product that was commercial, that was practical enough to be important in a healthcare setting.”

The Five-Year Customer

Healthcare doesn’t move fast. Five years ago, Henry presented to a healthcare company’s board. They believed in it. Then nothing happened for years.

“We’re working with one of the organizations, healthcare companies, that five years ago was the first introduction. And I spoke to their board of directors nearly five years ago, presented what we were doing. And the truth is, even then, there was a belief on their part that what we were doing was both important and quite real.”

Henry stayed engaged. “I stayed connected with this healthcare institution because frankly, I liked the people, I trusted them.”

The payoff was critical input. “They also gave us insight into how it might impact in the organization. How would you bring it in so that it was positive and it augmented this interaction between patient and doctor.”

Three years ago, that institution invested. Then they commercialized. That relationship unlocked Microsoft partnerships worth $1.5 million in non-dilutive grants.

What They Built

Today, Canary Speech analyzes 15.5 million data elements per minute from 42 seconds of speech. They detect mild cognitive impairment with 87% accuracy and Alzheimer’s with 93-96% accuracy. They work in multiple languages across multiple countries.

“If a doctor is doing analysis for, say, Parkinson’s disease, while they’re speaking with the patient on their phone, they’re seeing a measurement come back of the individual’s assessment for Parkinson’s. And in addition to that, they’re seeing an assessment for their anxiety, their depression, and their fatigue all in the same 40 seconds,” Henry explains.

The information is objective and arrives in real time.

Building the Future

The technology works anywhere there’s a smart device. “It could be rurally deployed. It can be deployed anywhere there’s a smart device, like a smartphone or a tablet or through video conferencing.”

Partnerships are expanding—Microsoft and Nuance integration, healthcare institutions across continents, work from Japan to Ireland to the Emirates.

The vision is about persistence. “You’ve got to put the stake in the ground, and that stake may be three years out, it may be eight years out, it may be ten years out.”

For Canary Speech, that stake was planted in a bagel shop with a question: what if we could detect disease not by what people say, but by how they say it?

“I believe individuals that are inspired to work together can do things that are considered impossible. And as the saying goes, it simply takes a little longer.”

Eight years. Twelve patents. $26 million raised. Technology detecting a dozen diseases from 40 seconds of conversation.

“The journey that you take to where you get that sounds ridiculous, but you never know. You just don’t,” Henry says. From bagel shop to building the future of voice-based disease detection—that’s Canary Speech.