Beyond the AI scribe: How artificial intelligence is transforming healthcare from drug discovery to patient care

The widespread adoption of AI scribes across the country has contributed to alleviating the administrative burden on doctors and provided tangible evidence of AI’s capabilities, but AI scribes are only one example of the technology’s impact on healthcare.AI is also impacting drug development, population health analysis, personalized medicine, radiology and even organ allocation. It’s pervasive, not only in healthcare but in virtually every sector of the economy, including agriculture, mining and manufacturing. It’s on our phones, it’s in self-driving cars and trucks, it’s everywhere.A panel on how AI is revolutionizing patient care and medical innovation at the ALL IN 2025 conference, Canada’s premier AI conference held in Montreal September 24 and 25, offered an overview of AI’s impact on healthcare beyond AI scribes.ImagingRoxana Sultan, chief data officer and vice-president of health at the Vector Institute, one of three AI institutes established under the pan-Canadian AI strategy, highlighted AI’s impact on medical imaging and its use at Toronto’s University Health Network analyzing chest X-rays to detect cases of pneumothorax.“The model we trained using thousands of historical chest X-rays is now achieving a 96% accuracy rate and can detect even the most subtle patterns that might be missed by the human eye,” said Sultan. “This allows us to prioritize the X-rays for review by a radiologist to ensure that patients most at risk are not left waiting too long.”Sultan also highlighted another model for analyzing MRI images and biomarkers such as the volume, size and shape of different parts of the brain. The model enables patient stratification and targeted therapies used for treating cognitive impairments associated with dementia and vascular disease.AI won’t “supplant that human capability provided by expert clinicians,” but “can augment and support decision making,” she added.Organ allocationAI was also instrumental in remedying bias and inequity in organ allocation at Toronto’s University Health Network where “researchers found that women were systematically disadvantaged by standard organ allocation processes for liver transplants because their symptoms and conditions were not identified as being as severe as they actually are,” said Sultan. “Using deep learning techniques on complex datasets,” she said, “a model was created that could predict outcomes much more efficiently and effectively irrespective of a patient’s sex.”Read: Nova Scotia’s common EHR revolutionizes access to patient informationDrug discoveryDrug discovery, another important area of focus for AI, can be accelerated by literature reviews “to figure out what has been said about a drug, a disease or a patient population,” said Marc Bellemare, co-founder and chief scientific officer at Reliant AI. “We’ve completely surpassed how this was done manually,” he noted. “It’s a game changer because . . . AI can read thousands of papers for you when you can read one paper in the same amount of time.“In the last three years, we have gone from a state of ‘it’s cool and can do nice party tricks’ to ‘it’s better than us at a lot of things we can do,’” said Bellemare.AI can also speed drug discovery by improving the effectiveness and efficiency of randomized control trials. It does this, said Sultan, by analyzing electronic health data to identify candidates who would be most suitable for a trial and eliminating those who would be at most risk of dropping out, thereby “improving the likelihood of having a complete cohort at the other end of the research process.” It can also be used to monitor patients to address any adverse events that may arise during a trial.Dr. Rene Caissie, CEO and co-founder of Medeloop, highlighted the use of agents that can remotely access datasets wherever they reside without moving them. He cited the example of research at McGill University on Ehler-Danlos Syndrome (EDS), a disease characterized by overly flexible joints that can result in joint dislocations and early onset arthritis. McGill wouldn’t have enough EDS patients and data to build a model but could use agents to pull data from “multiple different silos and multiple different buckets without ever moving it,” said Dr. Caissie. The resulting AI model can be used to make new discoveries, find new biomarkers in hours instead of years and enable personalized treatment.Medeloop, which is based in San Francisco with an office in Montreal and plans to open another one in Toronto, is also partnering with Open AI to put together a dataset of throat cancer patients using the same agentic workflow. “We believe we can go from discovery to writing a JAMA or New England Journal of Medicine paper in an hour or two versus a full year,” he said.Dr. Caissie practised surgery for 10 years in Canada before relocating to Silicon Valley, where he attended the Stanford Graduate School of Business for training in AI.“Datasets in Canada are all in different buckets,” he noted. “They’re in the DSQ (Dossier Santé Québec), on Telus and Shoppers’ servers and at the government level. To make use of it, Medeloop deploys agents where the data resides.”In the U.S., said Dr. Caissie, Medeloop has access to 200 million patient records and will have access to 250 million before the end of the year.“When we think about the impact that AI will have on drug development, there are two complimentary axes,” he said. “There’s the speed at which we can process information and the volume of data we can ingest.”Medeloop worked with a top 10 pharma company, for example, to build a tailored database of epidemiology information for every disease in the world identifying how many people have the disease, and how many will contract it the following year. This research on a disease would normally be performed manually and take six weeks, said Dr. Caissie. “We’re able to do it in a day. It’s dramatically changing the way we work with information. It’s the same as when we have an argument at a restaurant and I can just pull out up my phone and settle it with Wikipedia. We couldn’t do that years ago, so this is dramatically accelerating drug development just by virtue of getting answers faster.”Dr. Caissie shared a personal story about the impact of AI on his own health, revealing he was diagnosed with cancer earlier this year. “I was treated with surgery and radiation and with my doctors we used AI to look at all of the literature. It changed the way I was treated and it’s going to change the way future generations are going to be treated. I’m cancer free today, so all is good.”Population health analyticsPopulation health analytics is another beneficiary of AI, noted Martin Belanger, senior vice-president of payer and provider solutions at Telus Health. Eighteen months ago, Nova Scotia Premier Tim Houston reached out with a request to unlock primary care data from the Telus Health EMR footprint in the province for inclusion in the YourHealthNS app, he said. In addition to providing Nova Scotians with access to their own health information, the resulting comprehensive dataset helped identify new programs for preventive health.Using Medeloop’s agents, Telus Health hopes to access other datasets to create actionable insights and identify care gaps across its portfolio.The ALL IN 2026 conference on AI is scheduled for Sept. 16 and 17, 2026 in Montreal and will be expanded to also encompass quantum computing and cybersecurity.

Beyond the AI scribe: How artificial intelligence is transforming healthcare from drug discovery to patient care

Register for free access or login.

Related Topics