AI has no shortage of promise—or predictions. When generative AI first emerged, many expected it would completely transform everything, and stories about what frontier models can do keep those expectations high. But in healthcare, I see a disconnect between what people believe the technology is capable of and the everyday reality at the point of care.

Make no mistake: AI is very capable. There is simply a significant amount of complexity between having a powerful piece of technology and translating it into real impact—deploying it safely, testing and evaluating it, making it transparent, earning user trust and fitting it into existing workflows without being disruptive. In my experience, closing that gap comes down to two things: trust and ownership.

Stepping out of the shadows

One sign we are still living through a kind of “Wild West” period for AI is the prevalence of shadow AI. The term can sound nefarious, but it really refers to people using AI in ways that aren’t visible, approved or incorporated into an organization’s established governance and procurement processes. A physician might bring their own ambient listening solution into a practice without health system approval, or clinicians might use general-purpose chatbots to summarize information or assist with patient assessments.

The implications are real. There are many regulations around AI in healthcare, and various states require disclosure of AI use to patients along with consent and opt-in or opt-out provisions. Ultimately, the responsibility falls on the health system. When providers use AI in ways that are out of compliance with internal policies—or when governance structures aren’t yet robust enough to communicate and enforce those policies—organizations carry risk they can’t see.

Questions every executive should ask

When I speak with healthcare executives about AI adoption, my first piece of advice is to recognize that every AI initiative affects multiple stakeholders. With financial pressures top of mind for nearly every leader today, you need to understand how a proposed initiative affects patients, clinicians, operational teams and organizational finances—whether you’re a clinical leader, nursing leader, finance executive or CEO. When you understand how your projects impact different parts of the organization and the things they care most about, you can cultivate better alignment and “buy in”, which helps drive adoption and willingness to learn and adapt. It’s critical that every person across an organization knows exactly how AI can work for them and bring positive results to their teams.

When evaluating vendors, look for true partners who give you a degree of control. Providers are still responsible for patient care decisions, and AI should never remove that responsibility. The best AI partnerships enable health systems to configure solutions, incorporate their own clinical guidance and maintain oversight—and providers should be able to understand how AI arrived at a recommendation and review the evidence behind it.

Finally, be cautious of vendors claiming they have a fully baked, nearly perfect solution. When someone says a tool is 99.9% accurate and essentially bulletproof, ask questions: What population was it trained on? What population was it tested on? Will those results translate to your organization and your patients? Every health system is different, which is why the most effective approach is phased deployment. Start small, gather feedback, iterate and continuously improve.

AI adoption should be treated as a journey, not a one-time implementation—and the strongest outcomes happen when ownership is shared among providers, health systems and technology partners.

Where AI is already delivering

Ambient listening is the clearest success story in healthcare AI today, and the feedback has been remarkable. Clinicians tell us their lives have been transformed: They’re spending more time with their families and getting home at reasonable hours. We’ve heard of clinical leaders receiving sincere thanks from providers who feel they’ve gotten part of their lives back. Patients benefit, too, reporting stronger connections with providers who can maintain eye contact and be fully present during visits rather than concentrating on documentation.

One of the more surprising developments is that the benefits now extend beyond clinician satisfaction. Ambient listening often captures more detail and nuance than traditional note-taking, so documentation can more accurately reflect patient complexity and risk. That supports more complete charge capture and more accurate coding—real financial value on top of the human value.

The road ahead

Looking at the near-term horizon, interoperability is a major area to watch. We’ve talked about it for years, and AI now creates opportunities to connect community data, summarize it and make it truly usable for clinicians. Revenue cycle intelligence is another important category, with prior authorization and intelligent coding solutions becoming increasingly viable.

The technology is largely ready. The challenge now is diffusion: getting these capabilities into users’ hands, adapting workflows around them and building trust every step of the way. That’s a journey worth taking together.

To learn more about how Altera is using AI to bring the human side of healthcare even closer, visit our artificial intelligence page.