Four tips from healthcare leaders for finding ROI in AI

While AI is undoubtedly creating value in the healthcare industry, healthcare leaders are still struggling to quantify and capture it. Only about 54% of senior leaders report meaningful return on investment at their organization after their first year of generative AI implementations. 

At Noteworthy 2025, Notable’s annual summit on healthcare innovation, key industry leaders dove into this topic. Aida Mendelson, Solutions Lead at Notable, moderated a panel with Dr. Steven Packer, CEO Emeritus of Montage Health, along with McKinsey Partners Dr. Jim Kretlow and Dr. Sanjiv Baxi. 

Their discussion explored what effective AI adoption looks like and how organizations can structure their strategies to achieve measurable impact. Here are their tips: 

1. Build a business case that ties to real outcomes

Dr. Baxi laid out key principles for realizing AI ROI: focus and measurable value. 

“Have a business-backed perspective of what you’re trying to do… There are a lot of pain points in medicine. Many of them are not associated with value, and many of those are not associated with value you can measure. So don’t spend your time on things you can’t measure, and that don’t create value.”

“It seems simple,” he added, “but you would be surprised by how many health systems don’t do that.”

It’s also easy, Dr. Baxi warned, to spread resources thin with too many overlapping pilots. Fifteen pilots across 15 different topics is “probably not a good use of your time,” he stated. “It spreads everyone thin, and you’re going to dilute the value that you get in any one thing.”

His advice is to pick one domain, whether it’s revenue cycle, clinical operations, or contact center, and focus heavily there to create value before expanding. 

Adding to Dr. Baxi’s advice, Dr. Kretlow emphasized that clear success depends on disciplined planning tied to real outcomes. You need a clear idea of what opportunity you’re chasing, like reducing premium pay costs or improving length of stay, and then you need to tie that to what you expect AI will unlock. That becomes your business case. 

This will vary by use case, but to Dr. Kretlow, “Everything needs to track, to some level, to bottom line impact.”

AI scribes are a clear example of a great use of AI that hasn’t yet delivered full ROI. 

“We’re seeing a lot of deployment, but not as much translation into things like are we re-templating? Do we have shorter slots because of the time saved? Are our physicians more productive?” said Dr. Kretlow. “Without the clarity of a good business case and the operational changes downstream from this, you don’t capture the value that you should.”

2. Design solutions that solve the right problems 

Before jumping headfirst into an AI solution, it’s important to evaluate whether the roles AI is supporting are truly needed. 

Dr. Baxi urged healthcare leaders to consider, “Does that role actually create any value for healthcare, change the outcome for the patient, create value for the health system?”

If the only purpose of a role seems to be helping move along a broken system, he recommends considering moving forward with something else. “Fundamentally, at the end of the day, it’s about improving the outcomes for patients,” Dr. Baxi stated. If your AI solutions aren’t doing that, you won’t see real value. 

3. ROI isn’t always dollars — sometimes it’s dinner time

As a former CEO known for his focus on ROI, Dr. Packer added an important consideration: not all impact is captured on a balance sheet. 

Reflecting on the wide rollout of ambient listening tools, he said: 

“When we talk about return on investment, think about the pajama time we’ve reduced in all of our primary care doctors’ lives, in terms of getting them home at a reasonable hour to enjoy their families, maybe have dinner with their kids. These are hard to measure ROIs coming from a guy who is deeply interested in ROI… But they’re priceless in many ways.”

Hard ROI is crucial, but don’t overlook the harder-to-measure returns on provider or patient satisfaction. 

4. Don’t measure AI against superhuman standards

It’s important to remember that the goal of AI is to improve standards as they are now, not to perform at 100% immediately. 

Dr. Baxi stated, “It’s striking to me how we set a bar for technology and what we expect from it is far beyond any human.”

He gave the example of autonomous vehicles, which have driven 96 million miles across four US cities with a 91% reduction in serious accidents. Only one of 45 accidents recorded was the fault of the autonomous car. However, autonomous cars haven’t been adopted in every city in the US in response to these incredible results. 

“In any other field, a 91% reduction in a serious negative outcome would transform that business,” said Dr. Baxi. “It has not done so in technology because our bar is unrealistic. We often set it to be unachievable and superhuman. I’m not sure that’s the threshold we need to keep.”

Instead, compare results from AI to results when the same task is performed by humans. This will provide a clearer picture of ROI and build wider acceptance of AI across the enterprise. 

Committing to an AI strategy

Organizations that prioritize strategy are realizing AI’s true value in healthcare. 

As these healthcare leaders made clear, capturing ROI requires more than deploying the latest tools. It takes disciplined focus on measurable outcomes, alignment with real operational needs, and a willingness to rethink outdated processes and unrealistic benchmarks. 

Explore more insights from Noteworthy 2025 here.