Why self-service tools fail to serve the patients who need the most help, and how AI agents are closing the gap
Digital front doors, such as apps, portals, and scheduling tools, have become standard features in most healthcare systems. But less than half of patient healthcare tasks are completed through digital tools. For patients on Medicare or Medicaid, the rates drop lower. Where the self-service portal falls short, the phone call becomes the default channel.
This article outlines the key gaps we’ve identified: where digital self-service tools are practical, where they are not, and how AI agents are now stepping in to extend care team capacity without requiring additional hiring.
Self-Service Doesn’t Scale for Patients with Complex Needs
Digital self-service works when the task is simple, familiar, and low-consequence. But that is not how most high-need patients experience healthcare.
Across Medicare and Medicaid populations, adoption of portals remains modest. While most adults over 50 report access to a patient portal, fewer than 25 percent use one regularly. According to a 2019 HealthMine survey, only 30 percent of Medicare Advantage members found their portal to be easy to use. Nearly half weren’t sure if their plan even offered one.
For more complex navigation, such as arranging home health services, requesting referrals, or coordinating benefits, the drop-off is sharper. Only 30 percent of older adults said they would use a portal to request a referral, compared to 46 percent who preferred to call. For home health services and durable medical equipment, patient-initiated digital engagement is estimated at below 10 percent. These services typically require a provider referral and are managed through staff coordination or inbound phone support, rather than through apps.
Many Patients Can’t Use Digital Tools—Even When They Have Access
Patients most in need of follow-up, such as older adults, individuals with low incomes, and non-English speakers, are often the least likely to engage with portals, apps, or digital scheduling tools.
Among older adults, 78 percent report having access to a portal, but fewer than one in four use it consistently. Many Medicaid beneficiaries rely on prepaid phones with limited data, and broadband remains inconsistent in rural areas. Half of Medicare Advantage members with chronic conditions weren’t sure if their plan offered a portal, and only 30 percent found it easy to use.
These aren’t isolated issues. They’re patterns we’ve seen consistently across care management programs. When a patient needs to schedule a follow-up or verify benefits, they usually don’t log in. They call the front desk, or worse, wait for someone to call.
Only a Fraction of Tasks Can Be Handled Through Digital Self-Service
Even among the most digitally engaged populations, task completion through portals typically falls below 50 percent. And that’s for basic, recurring needs.
A recent Ipsos survey found:
41 percent of portal users refill prescriptions online
38 percent pay medical bills through digital platforms
34 percent use the portal to schedule appointments
25 percent use it to check insurance benefits
Fewer than 15 percent use it to coordinate post-acute services or manage referrals
When weighted by task complexity and population needs, we estimate that only 25-35% of healthcare tasks are realistically handled through digital self-service tools for Medicare and Medicaid populations.
That means 65 to 75 percent of healthcare navigation still relies on people. For the highest-need patients, that reliance is likely closer to 90 percent.
Call Centers Fill the Gap at Significant Cost
When the portal doesn't complete the task, the burden falls on the call center. We see this every day in our care management work. For context, we started Lena by providing one-to-one dedicated (human) care navigation for high-risk Medicare populations. We've since scaled across populations, geographies, and risk groups. We've learned that our telephonic navigators spend their time helping patients with care tasks like:
Scheduling follow-up appointments
Finding a specialist and booking an appointment
Navigating a referral to schedule an appointment
Calling insurance to understand and engage in benefits
Finding a ride to their appointment, getting groceries, or other social determinants of health resources
The real problem is that the tasks that self-service digital tools don't resolve require many human hours to coordinate telephonically with patients, providers, insurers, and community-based organizations. For health systems and care teams, this represents significant staffing costs, high management overhead, and varying quality results.
Where Digital Tools Fall Short, AI Agents Step In
When we examined how our in-house navigators spent their time, more than a third of their weekly hours were dedicated to calling referred patients solely to enroll them in services. Most of that time didn’t involve care. It was spent dialing, leaving voicemails, and waiting for someone to call back.
Calling patients takes multiple attempts. Even when someone answers, the conversation is usually the same. Patients often ask about the program, who referred them, and what happens next. These are not clinical questions, but answering them takes time. And that time pulls licensed staff away from the work that drives outcomes.
We studied the patterns from those calls and used that insight to design a telephonic AI that could handle the front end of the enrollment process. The agent, called Lena AI, places outbound calls, can retry unanswered attempts the following day, leave productive voicemails, and use texting to diversify outreach. We developed our own telephonic AI agent to address our workflow bottleneck.
Captions are auto generatedPlay
Now, our navigators receive patients who are already activated and ready for guidance. Our navigators couldn’t be happier. They spend more of their time building meaningful connections and helping patients move forward, exactly what brought them into this profession in the first place.
Beyond program enrollment, most of these routine calls follow a routine structure. We've developed telephonic AI agents that handle routine calls, such as wellness check-ins, appointment reminders, preventive screenings, and referral coordination, without requiring the involvement of a coordinator. When something more complex arises, the agent flags it and hands it over.
We also built agent "specialists" to manage escalations more effectively. These AI "specialists" are trained to gather additional information when a patient flags an issue. They follow structured protocols to ensure nothing is missed before passing the case to a care team member. Specialists are also designed to follow gold-standard protocols for specific use cases, such as triaging symptoms. Others are trained to answer frequently asked questions about coverage, scheduling, and program logistics so coordinators don’t have to.
This shift hasn’t just improved the day-to-day experience for staff. It has enabled us to scale caseloads and increase efficiency across programs, resulting in lower costs and expanded access for our health system and payer partners, who are under pressure to achieve more with fewer resources.
Closing the Loop
Digital front doors are here to stay. However, for patients with the most complex needs, they were never at the front line.
Most of the care journey still happens through phone calls, text messages, and real-time human support. AI agents, like Lena, don’t replace that reality. They make it scalable. They handle the repeatable work, structure the outreach, and bring access to care faster, without requiring them to navigate a portal.
If your team is facing similar challenges, reach out by messaging me directly - I'd be happy to share our experience or show you how Lena handles enrollment outreach.