By Aditya Bansod, Co-Founder and Chief Technology Officer, Luma Health
Over the past five years, health systems have invested heavily in the digital front door, and have even begun incorporating AI solutions like chatbots, aiming to address one of the biggest challenges in healthcare: an overwhelming administrative burden.
Yet most AI tools health systems are using still haven’t solved everyday inefficiencies:
- Chatbots direct patients to “find a doctor” pages, but they can’t actually book an appointment.
- Patients reply to text message appointment reminders asking to cancel or reschedule, but they’re prompted to call the office.
- Providers “click to refer” patients to a specialist, but the result is still a fax that takes weeks to process.
- AI call centers still require patients to transfer to a human agent.
These shortcomings are why the initial success with most AI implementations fades quickly. In fact, a recent MIT report found that as many as 95% of generative AI pilots at companies are failing to show meaningful ROI. It shouldn’t take 3-5 staff per provider to handle all the operational sludge behind the scenes, just to get a patient in the door.
Health systems are bleeding revenue, and patients are going elsewhere: patients report that they don’t recommend providers that are hard to get to.
The Reality of AI vs. Marketing Promises
Most AI tools fail because they’re essentially window dressing: they focus on the front-end, customer-facing interaction but ignore the back-end workflow. The digital front door invites patients in, but they immediately fall through the floor.
Because they don’t operate deeply enough within the EHR, these front-door solutions create a false promise — the experience looks slick but isn’t actionable. The AI can’t actually perform the tasks patients and providers need.
Health systems took action to implement pre-agentic AI to stay ahead of the curve, but many are now seeing:
- Continued bottlenecks and inefficient workflows.
- Double-documentation and redundant tasks across multiple systems.
- More time, frustration, steps and work for staff.
- Low conversion rates and unused capacity.
AI tools that fall short also hurt patients by delaying their care. Even with the best possible digital front door, nearly half of consumers say they still encounter friction, and almost a quarter give up on booking.
You can’t cancel a subscription today without being routed through a customer retention workflow. So why are we allowing patients needing healthcare to fall off the radar?
AI That Actually Does the Work
Health systems don’t need another wayfinder or receptionist. They need an operational plumber — AI solutions that actually work behind the scenes to connect workflows and platforms to streamline processes and the patient experience.
Unlike superficial systems that only serve as a front desk, FAQ or bolt-on automation, Luma's operational AI functions inside the system to move patients from Point A to Point B, and solves the biggest healthcare hurdles:
- Accelerated referrals: With more than 1,500 incoming faxes per day, the staff at DENT Neurologic Institute couldn’t process referrals fast enough. Patients in pain waited weeks for appointments. Luma’s Fax Transform processes faxes and automatically takes action by prompting referred patients to self-schedule. Referrals are processed same day instead of in weeks, and 15X more faxes get action daily.
- Real-time appointment management: With the call center drowning each morning in follow-up from after-hours calls, the University of Arkansas Medical Center couldn’t afford another workflow that funneled patients to the phone lines. Luma’s Navigator conversational AI helps patients manage their appointments after hours, with writeback to Epic. UAMS automates 95% of inbound calls and saves over 800 hours annually on after-hours calls alone with Luma’s AI agent handling these needs.
- Provider-initiated cancellations: Managing the schedule and calling every patient to reschedule when a provider is unexpectedly out can derail an entire workday. At a large Epic-first health system in Pennsylvania with 500+ of these appointments per day, Luma completely handles the process — from finding providers who are out, to reaching patients individually to reschedule, to writing back the changes directly to Epic. Within months, the system saved over $1M in staff time.
Why Operational AI, Not Surface-Level Automation, Is the Answer
With patient expectations and labor pressures rising, health systems are demanding more from their AI tools. Every point of friction — the cancelled appointment that’s not rebooked; the referral that isn’t processed quickly — has real financial cost that no one can afford.
The era of AI for AI’s sake — the first circa-2024 wave of AI without deeply EHR-integrated agentic capabilities — is over. Health systems don’t need flashier front doors. They need outcomes. Fully integrated AI solutions that do the work behind the scenes can plug revenue leaks and deliver the efficiency, time and cost savings that staff and patients expect.