TORONTO — Signal 1, which makes AI tools for health care, has landed a deal with Newfoundland and Labrador’s medical authority as it grows a Canadian clientele that includes Nova Scotia’s provincial system and hospital groups in British Columbia, Ontario and Quebec.

The Toronto-based startup sells an AI management system (AIMS) that customers use to evaluate new applications and agents, as well as track their use and performance. The platform can also calculate the costs of the tools, measure their impact and warn of any problems they generate.

Health care is “still very labour-intensive,” said Signal 1 CEO Tomi Poutanen. Hospitals and medical centres are increasingly adopting AI agents for administrative tasks that take up a lot of staff time but don’t directly treat the sick, like appointment and shift scheduling, handoffs between clinics and follow-up calls for discharged patients. That, Poutanen said, has created the market for AIMS.

Poutanen, the former chief AI officer at TD Bank, founded Signal 1 with University of Toronto professor Mara Lederman in April 2022. The firm originally set out to commercialize medical AI models developed at St. Michael’s Hospital in Toronto, but shifted to selling AIMS in response to demand from U.S. clients.

Signal 1’s clients across the border include major hospital networks such as Inova Health and Mount Sinai Health System, as well as NewYork-Presbyterian, announced last month. U.S. customers are typically running more than 200 AI tools and applications, according to Poutanen. “If you’re only managing a small handful of AI solutions, you can do that on the side of your desk,” he said, but as the number climbs, “that becomes untenable.”

In Canada—where organizations have been slow to adopt new technology across most industries in general and health care in particular—most individual hospitals are running 20 AI applications at most, and don’t have much money for innovation. So Signal 1 is focusing at the provincial level, where health authorities are tasked with “delivering universal, high-quality care at the lowest possible cost to the taxpayer,” Poutanen said. “Their incentives are to innovate.”

Provincial deployments often cover the same number of hospitals and patients as the larger U.S. health systems, Lederman said. And since provincial governments ultimately pay the bills for health care in Canada, they can see the direct benefits of AI in boosting staff productivity, or reducing the cost and increasing the quality of treatment.

The Nova Scotia Health Authority has adopted or is considering AI tools that predict nurse staffing requirements, help patients better navigate the medical system, surface clinical data to doctors faster, and draft reports based on medical imaging. The agency has also bought access to a Health Canada-certified application from Signal 1 for patient monitoring, said chief information officer Scott McKenna. 

Toronto’s North York General Hospital is using a Signal 1 algorithm to help optimize its emergency department so patients receive treatment faster, and fewer leave before being seen. The project helped show “the value that AI can bring to us,” said chief digital officer Duska Kennedy. It was also a way to test Signal 1, which won the hospital’s ongoing business by sending staff in to watch how it operated in reality and working with it to design the final application. 

Neither Nova Scotia Health nor North York General are yet using AIMS, but both are planning to roll it out. The provincial authority wants to use the platform to help decide which AI applications to adopt, and to monitor the ones it does deploy. “If we’re trying to compare something in diagnostic imaging with something in cardiology, it’s apples and oranges,” said McKenna. He hopes AIMS will provide a standardized way to assess the value and risks of different tools. 

Nova Scotia Health will also use the system to ensure AI applications are only being used as intended, keep within safety guardrails and aren’t duplicating what other tools are already doing. “We can continue to evaluate every model and ensure that it’s still giving us the value that we need,” said Neil Phasey, the agency’s new chief AI officer.

Signal 1 has amassed lots of data and experience to help health-care systems understand which AI applications they can use for which tasks and to speed up adoption, Poutanen said. It’s built its own suite of over 300 evaluations for frontier and other AI models, tailored to health care. For example, one might measure the time it takes a tool to intervene when an alert goes off. “It’s a very heavy effort in automating as much of this process as possible,” he said. “By virtue of us being health-care-specific, we come to a customer with a point of view of what good AI looks like.”

North York General would ideally like to be using AI features that are already built into its existing core software systems for patient records, health information and human resources, according to Kennedy. Since those aren’t yet available, the organization has instead bought tools from several different vendors. Signal 1’s technology can pull information from each of those systems, providing “a more seamless and holistic way to evaluate all AI work within a hospital,” said Kennedy.

That ability to integrate across systems for health information, administration and customer management ensures Signal 1 won’t be displaced if any of those developers add AI monitoring tools, Poutanen said. And while many major software firms now offer AI “control towers,” the firm’s focus on health care gives it an edge. 

“In other industries, you can let AI and agents run autonomously,” he said. “In health care, it is a matter of getting the current clinical and the operational staff to work together with agents.” Plus, medical clients need to monitor patient outcomes, not just financial returns. While agentic control systems are increasingly in demand, Signal 1 has no plans to expand to other sectors.

Other Canadian clients include Montreal’s Jewish General Hospital, Toronto’s Unity Health, Vancouver Coastal Health and Trillium Health Partners in Mississauga. Lederman said it has held early discussions with at least one other provincial health system in Western Canada.

The firm charges subscription fees for its tools, typically on three-year contracts. It could eventually look to take a cut of clients’ AI spending, for example, when it reduces their bills for tokens—the measuring unit of model outputs.

While U.S. health-care systems are adopting AI faster and more widely than Canadian peers, Lederman sees an opportunity to use Signal 1’s experience across the border to help hospitals here catch up. Canadian health-care leaders also need to set bigger goals for the technology, and governments need to fund them to scale it up across their operations.

“Our hospitals have to move beyond one solution for one little metric for one little problem, and start thinking about AI as an enabling technology,” she said.