At the recent HIMSS AI in Healthcare Forum, an interesting paradox emerged. Despite excitement about ambient scribing and AI agents, health systems are still spending more time governing AI than actually deploying it.
When a new clinical AI scribe agent made its debut, the room buzzed with excitement. But the very next session focused on oversight, governance, and policies; the same bureaucratic theater that seems to unduly slow healthcare innovation and transformation.
Implementation of operational AI solutions is gaining wide acceptance. According to a 2023 Health Management Academy survey, about 75% of large U.S. health systems reported they are either using or actively planning to use AI in operational areas.
At the same time, valid concerns remain, centered on ethics and accuracy, patient trust, organizational readiness and more.
As both a physician and a health tech executive who works to help health systems realize AI’s benefits to call center staff and patients, I witness this dynamic daily. While health systems spend 18 months developing perfect AI frameworks, more nimble competitors are already testing, iterating, and scaling solutions through real-world deployments.
While many caution against moving too fast with AI, we may be ignoring a greater risk : moving too slow while patient needs go unmet and operational challenges compound.
So how should health systems best move forward responsibly?
Here’s an approach I’ve seen work repeatedly:
Testing and Validation
- Maximize Impact: While you don’t have to start out with a big bang agentic AI launch, you do need to choose a problem to solve that is large and impactful. The value in deployment of an AI solution is based on scale. You need to have a high volume of transactions to truly appreciate the magnitude of the ROI.
- Crawl, walk, run: Just because you choose a big, audacious, problem to solve doesn’t mean you have to go all in on day one. As long as the problem you are solving has sufficient scale, you can work with vendors to develop an approach to ramping up the volume of interactions for your AI solution. This way, you can gain confidence and balance risk and reward.
- Be nimble: Often, organizations don’t start down the path to AI and automation due to concerns about being locked into expensive technology that may or may not deliver. Be sure to implement solutions with proven results, a solid track record, measurable ROI and the ability to tailor based on your needs.
- Looking for perfection is a recipe for failure: Humans are the most fallible technology health systems deploy. Yet time and time again, I interact with health system executives who expect perfection of AI solutions. Humans have biases and bad days, make mistakes, do things you trained them not to do, and are inconsistent. AI agents aren’t perfect either, but they are more consistent, resilient, patient, multi-lingual and accurate for most administrative tasks.
Change Management
- Start with administrative solutions: Change in an organization is challenging. But there is less friction on the administrative side of the house than on the clinical side. Upside for improved patient experience, enhanced revenue and reduced costs is high, while the risks are lower.
- Keep patient outcomes at the forefront: Case in point: A health system where the same call center agents who were answering a flood of low-complexity calls also were tasked with coordinating life-saving organ transplants. Patient care was clearly jeopardized in this arrangement. Implementing an AI-powered communications platform put human-centered care back at the forefront, allowing staff to focus on high-value tasks.
- Engage staff and evaluate regularly: Before implementing AI and automation, be sure to understand what’s fueling inefficiencies, causing staff stress and driving costs. Surveying frontline staff is the best way to understand daily challenges and encourage stakeholder buy-in. Evaluate by gathering staff feedback as well as data-driven analytics.
With health systems facing unprecedented challenges, it’s clear that the need for action is immediate. Following a prudent course in AI implementation doesn’t have to mean years of deliberation. Our health system clinicians, administrators and patients need less burden and better care today. As physicians and leaders, we must mirror the innovation of our counterparts in the business world. I am confident we can do this effectively and efficiently without compromising ethics or privacy.
About Adam Silverman
Adam Silverman, MD, is Chief Medical Officer at ActiumHealth. He is a passionate healthcare innovator with 30 years of experience as a physician and executive. He has a diverse history that includes private practice, academic medicine and physician leadership in both hospital and ambulatory settings. He is a board-certified General Internist. His experience includes Regional Vice President, Population Health at Saint Francis Healthcare Partners and Regional Vice President, Ambulatory Strategy and Continuing Care Chief Population Health Officer at Trinity Health in Connecticut.
