Responsible AI governance is helping Intermountain face the headwinds

The healthcare provider community is staring down some big changes over the next few years as the ongoing impacts of provider shortages and escalating patient needs collide with drastic reductions in federal funding and the gutting of public health infrastructure – all wrapped up in the rise of artificial intelligence: a category of technology that has already seen stunning successes and worrisome failures across the industry. 

Being an innovation leader at a large health system during this troubled time is no easy task, especially when that system is a non-profit that already runs on relatively small margins. But for Ryan Smith, Chief Digital and Innovation Officer at Intermountain Health, preparations for the gathering storm have unveiled a huge range of opportunities to reimagine what the future of AI-enabled healthcare could be. 

“There’s no denying that there are a lot of headwinds right now for health systems,” he acknowledged in a keynote address at the Trace3 Evolve Technology Conference, held in early October in Las Vegas. “We’re dealing with widespread burnout. We’re anticipating about a million new patients needing services in our geographical region. And over the next ten years, we’re going to have to take out somewhere between $350 and $400 million of annual cost just within our organization to offset the impact of the Big Beautiful Bill.” 

“That’s all going to be very challenging. But at the same time, I’m very optimistic that we can counter those problems with a sustained focus on innovative ways to simplify workflows and support both our caregivers and patients with AI. The key will be adopting an AI governance framework that prioritizes accountability, transparency, and reliability to put the right guardrails in place.  

A blueprint for responsible AI governance   

Like the vast majority of health systems, Intermountain is constantly evaluating how to infuse AI into the clinical and operational ecosystems.  With hundreds of potential vendor partners and thousands of ideas pouring in from both internal and external sources, it’s easy to get overwhelmed – or to get trigger happy with implementing solutions without a structured and repeatable approach. 

Intermountain stays organized by starting with a questionnaire to conduct a preliminary assessment. If the solution passes first inspection, the system then conducts a risk assessment and mitigation evaluation. Depending on those results, it may be handed over to the AI Governance Council for the next steps toward implementation. 

All of these assessments are rooted in the seven principles of robust governance: accountability, transparency, reliability, ethics, equity, privacy, and security, Smith explained.  A series of leadership panels, including executive leadership, technology experts, governance gurus, and a dedicated AI subcommittee, collaborate on ensuring that all solutions – whether brought in through partnerships or home-grown – pass through this rigorous governance framework. 

“We say ‘no’ to an awful lot of AI, because it doesn’t meet these criteria.  That being said, I’m really proud that we have deployed into production over 300 different AI tools to date.  So we’re not afraid of AI – we’re actually very bullish on it – but we’re also not afraid to pass on something if it doesn’t fit into our governance framework.”  

“We’ve coalesced the formula down to five key principles, which are to lead with operations, utilize the risk framework, take a varied approach to partnering, buying, and building, leverage our differentiated assets, and engage our caregivers throughout the process,” Smith said. “This is our secret sauce for success.”  

Putting principles into action with an AI solution to a common problem

This governance-led approach to AI implementation showed its mettle during Intermountain Health’s recent EMR consolidation, when the system migrated from eight disparate systems down to a single instance of hosted Epic. Leaders anticipated upwards of 50,000 IT help desk tickets to come pouring in immediately after the transition, and wanted to ensure that the organization could handle the massive influx appropriately. 

“I don’t care how many IT people, informaticists, or third-party contractors you have, it’s always going to be overwhelming,” Smith said. “But it can be a lot less overwhelming with the right AI tools to help out. In our case, that’s an AI triage agent that became the first point of contact for users experiencing issues after the rollout.” 

Instead of sitting on hold with a backed-up IT department, the AI agent allows users to pull up a dashboard within Epic and submit a description of the problem they’re having. The agent either provides an answer immediately, which avoids the need for a call to a human, or uses historical data to triage the issue and place it in the right queue for the IT team to answer. The model has even been tuned to know when additional information is required from the user, and can ask follow-up questions to help fine-tune its decision. 

“As a result, any tickets that do come into us are very close to 100% complete with the information we need to address the issue,” said Smith. “We’re not having to call users back and try to get them on the phone to gather more info, which can be very difficult. And we’re not having to hunt through a haystack of lower-priority tickets for the few critical things that could really impact care delivery. Instead, we can be fairly confident that the high priority tickets are being surfaced quickly so we can get them taken care of.” 

Smith was candid about the fact that the AI agent wasn’t perfect from day one.  

“We were probably about 60% accurate when we started. But by the middle of week two, we had the model up to around 90% accuracy, which was much better than the routing of our own service desk.  We’re pretty proud of those results.  And more importantly, our users love it because they don’t have to sit on hold and talk to service desk agents. The resolution process is much simpler and much faster.” 

Addressing the future with optimism and innovation

Despite already having hundreds of AI tools in various stages of deployment, Intermountain Health isn’t planning on slowing down with its pursuit of more efficient and effective ways to reimagine workflows. 

Upcoming projects include a care intelligence platform to equip caregivers with clinical insights, as well as expanded clinical decision support tools, diagnostic imaging capabilities, and even conversational outreach to provide a personalized AI assistant for every patient that touches the health system’s services.  

Some of these features may be developed in-house, but others will be the result of Intermountain’s strong focus on partnerships with new and upcoming AI solutions providers, as well as reliance on its existing core tech stack partners, to support the next phase of AI-driven innovation.  

“I love the African proverb that says, ‘if you want to go fast go alone. But if you want to go far, go together,’” Smith concluded. “Partnering our way to success is part of the long game we’re playing, because I truly believe we’re better together as we share ideas and develop new strategies to help people live their healthiest lives possible.”  

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at [email protected].