The World AI Cannes Festival is back Feb 8-10 and we're there! See live updates on what's new in AI this year
%402x.svg){kind=icon}
.svg)
Corti, a Copenhagen-based health tech startup, recently raised $60 million to expand its service: an AI co-pilot for healthcare providers. Assisting on nearly 55 million annual medical consultations in real time, Corti’s AI tool is designed to help triage patients, document patient interactions with administrative coding, provide analysis on those interactions, give “second opinions,” and provide training notes.
As a team deeply entrenched in AI initiatives for healthcare providers and payers, two particular aspects of Corti’s methods caught our attention: the approach of building AI models for healthcare without the direct expertise of medical professionals, and the quest to make the term AI “boring.”
While Corti’s cofounders have experience building productivity tools for medical professionals, their decision to building AI models without leaning heavily on the expertise of medical professionals has been controversial. Traditional wisdom dictates that domain expertise is crucial when developing specialized AI models.
Why deviate from this path? Corti CTO Lars Maaløe says they chose this approach because having a physician on payroll “could create bias in the system.” They train models on customers and outside researchers, not medical providers.
This approach is not without risks. Medicine is a field where the stakes are incredibly high. Misinterpretations or oversights by an AI system could have dire consequences. While AI can and should be used to enhance human expertise and identify novel correlations, completely sidelining medical professionals in the development process is an extremely bold and daring take.
Kevin McCall, Managing Director of AI, agrees. “Having medical professionals—open-minded, technically skilled, appropriately guided and supported professionals—involved with the creation and deployment of any AI technology is going to be the norm,” he maintains. “That doesn’t detract from the novel and powerful things that AI can do. On the contrary, it increases its relevance and impact.”
Certainly, we agree that AI can identify patterns and insights that might elude even the most experienced medical professionals. But, Kevin says, a tool like Corti’s needs to be “a collaboration, where AI performs rapid diagnostics and human experts stay in the loop to validate interpretations and curate the models for continuous improvement.”
In other words, you can get the best out of doctors and AI if you do it right. This is what AI augmenting and empowering humans is all about—embracing the innovative properties of AI while using domain experts to keep the system grounded in the realities of medical practice.
.webp)
In a technology context, calling something “boring” means that it’s become so ubiquitous and reliable that it's no longer considered novel. It's a testament to the technology's maturity and integration into everyday life, and that’s the dream for any groundbreaking tech. When electricity was first introduced, it was a marvel. Today, it's unremarkable—but try living without it.
For AI in healthcare, achieving this level of ubiquity and reliability is both a goal and a significant challenge. The healthcare industry has many complexities, from patient data privacy to the intricacies of medical diagnoses.
That means AI-centric solutions like Corti can’t afford to misstep. Similar to how autonomous driving accidents have slowed the deployment of and consumer trust in driverless cars, a few highly publicized false interpretations will erode the confidence of both medical professionals and patients.
The potential for dire health consequences has long been the battlefield between the pace of progress in tech and the need for careful testing in medicine. Corti needs to strike a balance of rapid innovation with intentional oversight and due process to avoid any unfortunate missteps and build sustained trust.
That trust appears to be on the upswing since the product went to market—Corti’s founders have seen a dramatic shift in medical professionals’ willingness to test out AI in their practices as ChatGPT and other large language models have become mainstream. This latest round of funding demonstrates the medical field’s burgeoning confidence in and increased appetite for integrating AI into their daily work.
If AI can achieve “boringness” in this space, it will mean it has seamlessly integrated into the healthcare workflow, providing consistent and reliable support to clinicians without causing disruptions. A boring AI is a successful AI, one that works behind the scenes to enhance healthcare delivery—not replacing current methods, but augmenting and streamlining them.
While using AI technologies like Corti might become routine, the impact they have on patient outcomes, clinician workload, and the future of healthcare is anything but mundane. A starting focus on AI as a means to ease the patient journey through ordinary healthcare interactions may be the key for providers to not only see quick returns on AI investments, but boost confidence in other applications of AI in the healthcare ecosystem.
Corti’s progress will be fascinating to monitor. If they can uncover unconventional correlations through patient data and outside research fed models, and prove their medical interpretations to be as accurate—or, potentially, more accurate—than traditional diagnostics, they could be instrumental to dramatically lowering the cost of healthcare.
%402x.svg){kind=icon}
Corti, a Copenhagen-based health tech startup, recently raised $60 million to expand its service: an AI co-pilot for healthcare providers. Assisting on nearly 55 million annual medical consultations in real time, Corti’s AI tool is designed to help triage patients, document patient interactions with administrative coding, provide analysis on those interactions, give “second opinions,” and provide training notes.
As a team deeply entrenched in AI initiatives for healthcare providers and payers, two particular aspects of Corti’s methods caught our attention: the approach of building AI models for healthcare without the direct expertise of medical professionals, and the quest to make the term AI “boring.”
While Corti’s cofounders have experience building productivity tools for medical professionals, their decision to building AI models without leaning heavily on the expertise of medical professionals has been controversial. Traditional wisdom dictates that domain expertise is crucial when developing specialized AI models.
Why deviate from this path? Corti CTO Lars Maaløe says they chose this approach because having a physician on payroll “could create bias in the system.” They train models on customers and outside researchers, not medical providers.
This approach is not without risks. Medicine is a field where the stakes are incredibly high. Misinterpretations or oversights by an AI system could have dire consequences. While AI can and should be used to enhance human expertise and identify novel correlations, completely sidelining medical professionals in the development process is an extremely bold and daring take.
Kevin McCall, Managing Director of AI, agrees. “Having medical professionals—open-minded, technically skilled, appropriately guided and supported professionals—involved with the creation and deployment of any AI technology is going to be the norm,” he maintains. “That doesn’t detract from the novel and powerful things that AI can do. On the contrary, it increases its relevance and impact.”
Certainly, we agree that AI can identify patterns and insights that might elude even the most experienced medical professionals. But, Kevin says, a tool like Corti’s needs to be “a collaboration, where AI performs rapid diagnostics and human experts stay in the loop to validate interpretations and curate the models for continuous improvement.”
In other words, you can get the best out of doctors and AI if you do it right. This is what AI augmenting and empowering humans is all about—embracing the innovative properties of AI while using domain experts to keep the system grounded in the realities of medical practice.
In a technology context, calling something “boring” means that it’s become so ubiquitous and reliable that it's no longer considered novel. It's a testament to the technology's maturity and integration into everyday life, and that’s the dream for any groundbreaking tech. When electricity was first introduced, it was a marvel. Today, it's unremarkable—but try living without it.
For AI in healthcare, achieving this level of ubiquity and reliability is both a goal and a significant challenge. The healthcare industry has many complexities, from patient data privacy to the intricacies of medical diagnoses.
That means AI-centric solutions like Corti can’t afford to misstep. Similar to how autonomous driving accidents have slowed the deployment of and consumer trust in driverless cars, a few highly publicized false interpretations will erode the confidence of both medical professionals and patients.
The potential for dire health consequences has long been the battlefield between the pace of progress in tech and the need for careful testing in medicine. Corti needs to strike a balance of rapid innovation with intentional oversight and due process to avoid any unfortunate missteps and build sustained trust.
That trust appears to be on the upswing since the product went to market—Corti’s founders have seen a dramatic shift in medical professionals’ willingness to test out AI in their practices as ChatGPT and other large language models have become mainstream. This latest round of funding demonstrates the medical field’s burgeoning confidence in and increased appetite for integrating AI into their daily work.
If AI can achieve “boringness” in this space, it will mean it has seamlessly integrated into the healthcare workflow, providing consistent and reliable support to clinicians without causing disruptions. A boring AI is a successful AI, one that works behind the scenes to enhance healthcare delivery—not replacing current methods, but augmenting and streamlining them.
While using AI technologies like Corti might become routine, the impact they have on patient outcomes, clinician workload, and the future of healthcare is anything but mundane. A starting focus on AI as a means to ease the patient journey through ordinary healthcare interactions may be the key for providers to not only see quick returns on AI investments, but boost confidence in other applications of AI in the healthcare ecosystem.
Corti’s progress will be fascinating to monitor. If they can uncover unconventional correlations through patient data and outside research fed models, and prove their medical interpretations to be as accurate—or, potentially, more accurate—than traditional diagnostics, they could be instrumental to dramatically lowering the cost of healthcare.
