The future of AI in healthcare: Insights from leading investors

Though their firms execute a range of AI investment strategies, the panelists were aligned on where AI is having early adoption success and what those trends indicate for future investment. Cain Brothers advisors agree that such optimism is well-founded given the traction AI is already having on healthcare workflows and its potential for deeper industry transformation.
 

The evolving definition of AI

At a time when almost every solution or offering is being labeled “AI,” the panelists pushed for a more rigorous definition.

“You can kind of sprinkle AI on anything, these days,” Marina Kusserow said. “It’s table stakes now for companies to be using generative AI. But AI is not just any workflow using machine learning. It has to be software that's really learning and adapting to real world use cases. So, for it to be true AI, we need to see continual learning and growth from the tool.”

Tod Pesses agreed. “We define it almost identically. It's software that adapts and learns. But we get really granular as to why it’s learning and adapting so quickly. Is it leveraging data? Is it because of where it sits within workflow? Does it have unique integrations that allow it to adapt and learn with each claim, for example.”

Asif Dhanani emphasized the importance of value creation: “From our perspective, having focused more on services, we look at whether the technology generates more value with more data or is it really just automation? There's a limit to the value that automation alone will create. If you’re not recursive and you’re not growing from it, it’s not adding value to your delivery.”

Early adopters

Providers — particularly hospitals — are leading the adoption charge, driven largely by necessity.

As Kusserow noted: “Hospitals are using AI as a survival tool in a lot of ways. Burnout is at an all-time high. Margins are super thin. AI can come in and offer a quick relief for operational issues where the pain is most visceral and where the ROI is cleanest and cheapest. You’d be hard pressed today to go to a notable hospital and not see them boasting about their ambient scribe solution. And a lot of adoption is just peer-to-peer sharing. There are tools being passed around physician groups with individual doctors saying, wow, this has really allowed me to spend more time with my patients.”

Payers have also been adopting AI. As Pesses said, “Payers are running more pilots just given the data and dollars that they have.”

Kusserow highlighted specific use cases gaining traction including prior authorization, utilization management, care navigation, patient access, and workforce augmentation.

Dhanani added that ROI is a pivotal driver. “If you’re not seeing the impact in your growth or your profitability, what’s your incentive to work with AI?”

Many organizations are concerned they will miss out on the leverage AI can potentially deliver if they don’t invest and implement it. That fear of missing out is driving a certain amount of adoption.

The trust factor

How much is trust still a barrier to AI adoption?

On the clinical side, Pesses mentioned that we are not likely to see AI prescribers any time soon given the existing trust gap, but there is adoption of solutions that provide capacity to nurses and clinicians. Kusserow believes provider trust is a precursor to patient trust. She elaborated that providers have been inundated with new technology in recent years, much of which hasn’t made their lives better or easier. They’ll need more transparency, clinical validation, and workflow fit to get on board and are asking questions about impact. “Is the AI really going to make my work easier? Is it going to fit into my existing systems? Or is it just going be adding more time onto my day?”

Dhanani believes the culture of the organization plays an important role in their level of trust in the tool and openness to adoption. More agile and innovative organizations understand that no implementation is perfect from the beginning and benefits accrue over time. Adoption is also faster when vendors are a real partner with their customers in working through specific use cases. Few healthcare organizations have the depth of AI knowledge needed to make AI work on their own.

Pesses added, “It goes back to the importance of having these tools deeply integrated in your systems across specific workflows while utilizing the right data in the right way.” 

Making the sale

Convincing internal decision makers to purchase AI has become easier. According to Pesses, “Eighteen to twenty-four months ago, there were a lot of headwinds in the sales cycles, but I think customers know more of what they’re looking for today.”

However, budget allocation remains complex. Kusserow explained that customers are redistributing their budgets to make their AI purchasing decisions. “Not a lot of hospitals have the ability to just create a net new line item. So, they’re dipping into budgets that usually go to IT and quality management and things of that nature. We’re even seeing AI companies going after workforce budgets and saying, you’re having a hard time finding those 100 nurses because the labor market is getting smaller. What if we reallocate a portion of that budget to these AI co-pilot agents and devices?”

Pesses and Dhanani have noticed organizations taking a committee approach to such decisions. As Pesses said, “You definitely see the CIO more involved, but you also have department heads looking at how AI will drive workflows and where the ROI will be. To make the sale now, you have to evolve your messaging and sometimes your packaging to a different audience.”

Kusserow added that, “Every hospital executive we talk to, especially the CTO, says ‘you’ve got to be better than Epic because it only makes sense for us to try out the Epic product that is already embedded.’ That really pushes our startups to be better than the legacy solution that is such a force in our industry.”

Looking ahead

As the discussion concluded, the investors reflected on which metrics will define AI success in healthcare. Kusserow said the emphasis should shift from efficiency to human impact. “How will doctors be spending their time? Are they more excited about their work?”

Dhanani agreed, “I think in five years, we’ll be talking more about the outcomes we’re actually delivering for the healthcare ecosystem and healthcare consumers. It will be less, ‘Are you using AI?’ And more, ‘What are you doing with AI and what outcomes are you achieving?’”

Pesses added that we may not know exactly where AI will be in the next five years, but the future looks bright. “We’re incredibly excited about the opportunity to deploy capital and support entrepreneurs who are building really impactful businesses serving payers, providers, and manufacturers.”

Cain Brothers experts are also optimistic about AI’s five-year outlook. While AI value creation is in its early stages, and perhaps more aspirational still than real, we believe that over the next half decade, AI deployment will drive significant operational and administrative efficiencies. More importantly, AI will also begin to transform healthcare delivery itself while supporting better clinical decision-making and deeper patient engagement.

A podcast recording of the full conversation accompanies this article.