What we heard and why we went to listen
Last week, some of my CBTS colleagues joined hospital and health system leaders in Chicago for Becker’s 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health. We went to listen, learn, and connect so that we can better support healthcare clients as they navigate the opportunities and risks inherent in AI and digital innovations.
Here’s what we heard and observed and how it’s shaping our approach to supporting hospitals, health systems, and other healthcare organizations.
Buy, build, or co-manage decisions are top of mind
One session brought together a data leader, applications leader, and program manager from a large academic medical center’s language-model team. They walked through how they decide when to buy technology, when to co-manage it with a partner, and when to build it themselves.
Two things stood out. First, the deciding factor is usually the cost of operating it in year two and year five versus the cost of acquiring or building it in the first place. Second, these teams have largely accepted that they can’t staff an internal group for everything, so they buy where they can and reserve internal build effort for workflows specific to their organization. These are the workflows that require institutional knowledge that an outside partner simply can’t offer.
Vendor sprawl has become its own operational problem
Both on stage and off, healthcare leaders made it clear that they’re overwhelmed by the sheer volume of vendors calling on them. They described being approached constantly, often by companies offering something narrow and presenting it as significant. We sensed an appetite for providers that can deliver more comprehensive support rather than a single shiny object sold as a solution.
Vendor sprawl is also a burden in day-to-day operations. When internal teams are stretched thin, systems lean harder on outside providers. But it can be a challenge to stay clear on which provider owns what. Consolidating tools and reducing the number of relationships to manage came up as a goal, not a nice-to-have.
Guardrails and liability are setting the pace
Every AI conversation ran into HIPAA and protected health information within a few minutes. A physician we spoke with made the point directly: No matter how much efficiency AI delivers in a clinical workflow, the diagnostic risk still sits with the provider.
That explains where the near-term work is gaining traction. Several systems described defensible wins in revenue cycle management, where the operational lift of coding and billing is real and the clinical risk is contained. Automation is absorbing routine claims work, which frees experienced staff for payer behavior analysis and upstream problem solving. Leaders were candid that the most durable results are going to organizations reskilling their people alongside the technology.
More ambitious work is underway, as well. Examples include chart review to surface possible diagnoses and pattern analysis across historical records to flag patients at risk before a condition becomes serious. The sequencing is intentional and will be determined based on risk.
Governance isn’t keeping pace with deployment
The need for AI training and literacy was a recurring topic of discussion as organizations face the need to redesign entrenched ways of working. Yet no word came up more than governance. What leaders described was a widening gap between how quickly they can deploy AI tools and how well they can evaluate, monitor, and course-correct them. Successful organizations have built frameworks clear enough to scale a technology on evidence and to shut one down when the evidence isn’t there.
Leaders conceded that the act of shutting down can be a challenge. In one session, a speaker argued that every AI pilot needs an expiration date. Without a defined decision point and a success threshold set before launch, pilots become permanent fixtures that consume bandwidth and staff attention without ever being judged on their merits.
Problems that predate AI are still here
This may be the most important thing we heard because it got less stage time than it deserved.
In one session, the subject was the day-to-day reality of running a security operations center: 24x7 monitoring, keeping penetration testing on a sensible cadence, and covering it all with internal staff. In another, leaders talked about turnover and the high costs of hiring and training a specialist, only to absorb the loss when they leave a few months later. It surfaced a crucial question: At what point is bringing in a managed services partner the more responsible answer?
None of that is an “AI problem,” but it all determines whether an organization’s AI work has anything solid to stand on.
Tell us: Where does it hurt?
We support health systems today across networking, security, and communications work, and we’ve been building and operating technology foundations for more than three decades. That gives us real confidence in our engineering depth. It does not make us experts in your hospital or health system. That’s why our first question isn’t about technology. It’s the same one your clinicians ask every day: Where does it hurt?
So we aren’t going to tell you what your organization is doing wrong and offer one capability as the answer to all of it. (You’ve no doubt heard that pitch many times this year already.) What we would rather do is bring questions and a set of capabilities, understand which ones fit, and start with something targeted. As you well know, a hospital doesn’t have the luxury of shutting down for a technology cutover. Smaller, sequenced improvements are more realistic and sustainable.
Let’s stay in touch
If you were at the Becker’s conference and heard something different — or if the pressures in your organization look different from what we’ve described here — we want to hear it. If you’re planning to attend the Midwest Healthcare Innovation Summit 2026 hosted by Bamberg Health, be sure to look for our CTO Mark Giles. He’ll be part of a panel entitled From Data to Discipline: AI, Analytics & Enterprise Execution on Wednesday, September 30.