There has been a huge demand for healthcare AI this year — with three-quarters of the country’s providers and payers having increased their IT spending over the past year. This enthusiasm for AI in the healthcare sector has given way to hundreds of startups selling AI-powered products.
Although healthcare is a massive industry, its markets are fragmented. Because of this, AI startups should be thoughtful about which healthcare market(s) they’re going after. This is a factor investors pay a lot of attention to, according to a new report from Bessemer Venture Partners.
“The healthcare ‘system’ is a reflection of its evolution, which has been a gradual, piecemeal process rather than the result of comprehensive planning. For instance, the rise of employer-sponsored health insurance in the mid-20th century was an unplanned response to wage controls during World War II, while the introduction of Medicare and Medicaid in 1965 added new layers to the existing system rather than reorganizing it entirely,” explained Morgan Cheatham, vice president at Bessemer.
Solving Healthcare’s Provider Data Problem Starts with Interoperability
Break down the silos. Take control of your provider data.
These are just a couple examples of how incremental developments and responses to various pressures over time has led to a complex web of services, regulations and reimbursement, he noted. He also pointed out that healthcare is typically a local service, and this geographic variability further complicates things.
Healthcare AI startups seeking venture capital face heightened expectations for scale, which requires access to substantial total addressable markets (TAMs), Cheatham remarked. TAM refers to the total revenue opportunity available for a company’s product or service.
There are a few things startups can do to meet investors’ expectations. For instance, Cheatham recommended they target inherently large markets or consider operating across multiple markets.
He also said that a startup’s modality can have a sizable impact on its available TAM. By modality, he is referring to the way the AI product is delivered — some modalities include software, AI copilot assistants, diagnostic tools and therapeutics.
Navina’s AI Copilot Eases the Value-Based Burden, Study Shows
A new independent study from Phyx Primary Care’s Innovation Lab points to measurable improvements in physician burnout, documentation efficiency, and key VBC metrics among primary care practices.
Bessemer’s report noted that startups may want to think about combining multiple modalities within one comprehensive solution. For example, instead of just selling AI software, a startup could offer an AI assistant that helps doctors with their entire workflow to increase its TAM.
When AI models are deployed in just one isolated modality, it may be challenging for some providers to adopt them, the report said.
Cheatham also highlighted that there are sometimes trade-offs between achieving higher TAMs and having healthier gross margins.
To illustrate this concept, Bessemer’s report used a hypothetical startup in the ophthalmology space.
Say the startup sells AI software to ophthalmologists with a pay-per-seat mode — the company might have a smaller TAM because it is selling software to individual doctors, but it may also have higher margins because software has relatively low production costs. On the other hand, a company providing AI services for ocular injections might have a much larger TAM because they are selling a solution used in a common medical procedure — but their margins could be lower since services usually have more costs.
Photo: Andrey Tolkachev, Getty Images