- Healthcare startups increasingly want to use AI to care for more patients.
- Others are keeping the tech at arm's length over concerns about patient safety and experience.
- Here's how startups are upping caseloads, boosting margins, and hitting profitability with AI.
A growing number of healthcare startups are betting that artificial intelligence can help them do more with less — including, in some cases, to care for more patients.
The stakes are high: clinician shortages, tight margins, and mounting physician burnout are putting pressure on healthcare delivery businesses.
Startups say AI could be the breakthrough that helps clinicians multiply their caseloads and earn extra cash without sacrificing care quality.
"The long-term vision is to continue to peel away aspects of in-person care and deliver the care itself via technology," said CEO Daniel Perez, the cofounder and CEO of newly public physical therapy company Hinge Health, in a May interview with BI.
But while some companies are leaning all the way in to automate more and more care, others are keeping the tech at arm's length, wary of the safety risks or the patient experience of turning clinician tasks over to AI.
Instead, they're pushing AI behind the scenes to automate non-clinical tasks only. For some, that approach is offering a new lease on the startup's life.
"It's allowing companies that historically would not have gotten to profitability, that may have had to go out of business, to sustain and to be able to keep driving impact long term," said NOCD cofounder and CEO Stephen Smith.
More patients, more money
Some of healthcare's biggest AI moves are happening in musculoskeletal care, where companies like Hinge Health and Sword Health are aggressively using the tech to help make their physical therapists more efficient.
Sword Health spent 2024 ramping up its AI applications for clinicians. While its physical therapists were expected to manage between 200 to 300 patients at a given time at the beginning of the year, the company wanted those providers to manage 700 patients at a time on average by the end of 2024, BI reported in November. Sword declined to comment for this story.
Sword's efforts toward that goal included using AI to pre-write messages for clinicians to send along to patients, per BI's reporting. The company also said at the time that its AI helped physical therapists prioritize patients who might need more attention.
Hinge Health has taken a similar approach in implementing AI, including for patient-provider messaging. While Hinge Health declined to share its providers' caseloads for this story, the company says it's used AI to slash 95% of clinician hours spent on physical therapy.
Hinge Health also uses computer vision technology to guide patients through PT sessions at home. Following its May IPO, the company says it's actively working on more applications of AI to further automate care delivery, and has a large research and development team focused on that goal.
"At some point, whether 10, 50, or 200 years in the future, care delivery will be automated with technology. And that's a good thing," Hinge Health CEO Daniel Perez told BI in May.
Virtual pediatric care startup Summer Health is also working to use AI to multiply the number of patients its providers can care for. While founder and CEO Ellen DaSilva doesn't think AI will replace doctors, she said the startup is working on automating some clinical tasks as well as administrative ones.
Whereas a healthcare business with minimal technology implemented might expect providers to see five to seven patients an hour, Summer Health's largely text-messaging-based care model initially allowed its providers to see about 10 patients in an hour, DaSilva said. And with AI, "that number could pretty easily be doubled," she said. "Our providers still have a lot of room to run."
That could be especially good news for specialties like pediatric care where there's a shortage of clinicians. DaSilva said AI can help augment the remaining clinicians to help more patients get care, while hopefully reducing provider burnout along the way.
Balancing operations and outcomes
As some healthcare startups go all in on AI, other companies aren't quite as sold on its potential for improving care.
Chronic care company Omada Health, which went public in May, has long used AI behind the scenes, like to suggest educational videos in Omada's content library for clinicians to send to their patients. But it's been slower to bring AI in front of patients, waiting until the company could point to significant LLM improvements and train the models on its own clinical data.
The company released its first patient-facing AI tool earlier in May, an AI agent that answers members' nutrition questions. But Omada Health cofounder and CEO Sean Duffy said the company has no intention of meaningfully replacing patient-provider interactions across its business with AI.
"I've yet to find someone who feels accountable to Chat GPT," Duffy told BI in May at Omada's IPO. "I always quip that there's a reason that artificial intelligence is a buzzword and artificial empathy is not. You can expect us to always have a proactive people component."
Omada's diabetes care peer, Virta Health, has been using machine learning since 2017, per founder and CEO Sami Inkinen. Between 2015 and 2020, as Virta was figuring out how to use technology like machine learning to deliver better outcomes, it was turning a negative gross margin. But generative AI has helped the startup get that number in the green; today Virta boasts a roughly 60% gross margin, according to Inkinen.
Virta has injected AI into many parts of its business, including into patient interactions, with AI-powered personalized care plans and chatbots available to help patients answer nutrition questions 24/7. But Inkinen said AI isn't making clinical decisions and won't replace providers in Virta's business.
He added that Virta doesn't have caseload targets for its providers, and said the company is rigidly focused on delivering better patient outcomes and improving its financials along the way.
Mental health startup NOCD is using AI to tackle a broad range of administrative tasks, including clinical note dictation and revenue cycle management. CEO Stephen Smith said the tech has helped the company to reach profitability, allowing NOCD to grow its therapist network while keeping operating expenses low.
The startup does have a chatbot called Robin that therapists can ask for information during sessions. But that AI doesn't interact directly with patients, nor does it intervene unless the clinician seeks out its help.
Smith said he's treading carefully given the company's focus on patients with obsessive-compulsive disorder, which is typically classified as a serious mental illness. While NOCD may be able to increase its providers' caseloads with AI eventually, it's not a short-term goal for the company, he said.
"All these different projects, we know we have to proceed with some caution since we're dealing with a severe group of people," Smith said. "Bringing AI to the clinical delivery of care is going to require a lot of research, largely because we want to make sure it's super, super effective, and that we can create a safe experience for our members and our therapists."
