AI Is Everywhere in Health Systems. Access Is Still Broken.
Nearly every health system uses AI, and most patients still cannot get in. The AI went to the back office. Here is how to aim it at access.
Photo by Tim Mossholder on Pexels
Nearly every health system is now using AI, and most patients still cannot get in the door. That is not a contradiction, it is a clue. In my experience the AI went to the back office, where it saves staff time, and not to the front door, where a patient feels the friction. If you want AI to fix access, you have to point it at the access the patient actually experiences.
Two facts, from the same moment, that should not sit comfortably together. Managed Healthcare Executive reported that 91 percent of providers are now using AI tools, while 72 percent of patients still say they struggle to access care, and 51 percent have abandoned a care-seeking attempt because scheduling was too hard. AI adoption is nearly universal, and access is still broken. There is even a tell in the data: 56 percent of providers say they follow up with patients proactively, but only 25 percent of patients report actually receiving that outreach. And as Healthcare IT Today has documented, patients will avoid basic tasks like canceling entirely rather than make a phone call.
I do not read this as a contradiction. I read it as a map of where the AI went. It went to the back office, to documentation, coding, and revenue, where it genuinely saves staff time and money. It did not go to the front door, where the patient stands. If AI adoption is up and access is flat, the AI is simply pointed at the wrong problem for the person trying to get care.
The paradox in one chart
Put adoption and access on the same chart and the gap is the whole story: we are automating the parts patients never see. Ninety-one percent of providers using AI should have moved access. It did not, because usage and impact are different things, and the impact landed where staff feel it, not where patients do. The clearest tell is the follow-up gap: providers believe they are reaching out far more than patients report being reached. That is not lying, it is measuring the wrong end of the interaction. The patient is the one who has to feel the improvement, and by their own account, they mostly do not.
AI went to the back office
Most healthcare AI is aimed at internal efficiency, which is valuable and invisible to the patient, and that is exactly why access has not moved. Look at where AI actually gets deployed first. Ambient documentation helps the clinician. Coding and revenue-cycle tools help the back office. Prior-auth automation helps staff and sometimes speeds care. All of it is worth doing. None of it is the thing a patient experiences when they cannot get an appointment. The one row in the table below that the patient feels directly, self-service scheduling and reminders, is usually the last to get funded, not the first. Flip that order and the paradox starts to resolve.
I am not against back-office AI, I build it. But if your goal is access, measure access, and point the AI at the patient's actual friction: the phone call they will not make, the slot they cannot find, the reschedule that defeats them. The front door is where the patient meets your system, and it is where AI has barely shown up. Aim there, measure whether the patient felt it, and the adoption number will finally start to mean something to the person it is supposed to serve.
- 91 percent of providers use AI, yet 72 percent of patients still struggle to access care.
- 51 percent of patients have abandoned care-seeking over scheduling complexity.
- Providers believe they follow up far more than patients report receiving it (56 percent vs 25 percent).
- AI mostly went to the back office, where staff feel it, not the front door, where patients do.
- The patient-facing row, self-service scheduling and reminders, is usually funded last.
- If your goal is access, point AI at the patient's friction and measure whether the patient felt it.
Frequently asked
Why has AI adoption not improved patient access?
Because most AI is deployed in the back office, documentation, coding, revenue, where staff benefit, not at the front door where patients experience friction.
What does the data show?
91 percent of providers use AI while 72 percent of patients struggle to access care and 51 percent have abandoned care over scheduling, per Managed Healthcare Executive.
What is the follow-up gap?
Providers report proactive outreach at more than double the rate patients say they receive it, 56 percent versus 25 percent, a sign of measuring the wrong end.
How do I make AI improve access?
Aim it at patient-facing friction like self-service scheduling, reminders, and reschedules, and measure whether patients actually felt the improvement.
Is back-office AI a waste?
No. It is valuable, but it is invisible to patients, so it will not move access on its own. Access needs AI pointed at the patient.
Sources
Naveen Kumar
Healthcare engineering and product executive in Pittsburgh. 15+ years building AI-first patient access, a decade at Treatspace.