The Patient Services Automation Paradox
We’ve seen it time and time again: the push of automation where people want it least. But here’s what we should actually be building.
When was the last time you genuinely felt helped by a healthcare chatbot?
If you work in patient services, market access, or hub operations, you probably have your own answer to that question — and the answer probably involves a phone call you eventually had to make anyway. Here is the uncomfortable reality: the same industry that is living the frustration of inadequate patient-facing technology on a daily basis is also the one being pressured, quarter after quarter, to automate more, deploy faster, and do more with less.
The pressure is understandable. The healthcare AI market has surpassed $1 billion in value and is projected to reach roughly $10 billion over the next decade (PCG Software, 2025). Industry forecasts have suggested that as many as 90% of U.S. hospitals would leverage AI chatbots by 2025 to improve care and efficiency (Frost & Sullivan, as cited in PCG Software, 2025). Physician adoption of AI tools nearly doubled between 2023 and 2024, climbing from 38% to 66% in a single year (TempDev, 2025). The momentum is real, and no serious person in this industry is arguing against technology.
But momentum and direction are two different things.
The Catch-22 No One Is Talking About
A 2025 study published in SAGE Open Medicine surveyed more than 600 patients using a healthcare system chatbot and found that over 61% felt the chatbot did not fully understand them, while 35% reported they could not fully understand the chatbot (Shah et al., 2025). Those numbers reflect something most people working in patient support already sense: the technology has outpaced the foundation it is supposed to rest on.
The problem is not that patients reject technology. Research shows that roughly 50% of patients are expected to prefer using a chatbot for initial medical inquiries rather than waiting on hold with a human representative (PCG Software, 2025). Meanwhile, 60% of Americans reported feeling uncomfortable when their healthcare provider relies on AI for their care, while 64% said they would be comfortable interacting with an AI-powered assistant for basic questions or monitoring (TempDev, 2025). The nuance in those numbers is important: patients want technology to handle the things they find tedious, confusing, and time-consuming. They do not want it to replace the judgment, accountability, and human follow-through that determine whether they actually get on therapy.
That is a meaningful distinction. And the patient support industry has not fully reckoned with it.
Fast Is Only Valuable When the Answer Is Right
Nowhere is this tension more visible than in benefit verification. Electronic benefit verification, or eBV, has been marketed as a speed solution — a way to replace manual phone calls and faxes with near-instant digital responses. The capability is real, and in straightforward cases, eBV delivers on the promise. However, industry data indicates that in many cases, only 60% of electronic verification results are accurate, and more than 70% of prior authorization submissions are rejected with little to no reporting on the patient's status (EVERSANA, n.d.).
The problem is structural, not just technical. Payer data is notoriously fragmented and inconsistent, particularly for specialty and rare disease therapies that involve complex coverage rules, carve-out benefits, and frequent payer-level policy changes. Logic-based eBV solutions, which rely on pre-programmed coverage rules entered by humans, represent a static snapshot of the payer landscape — one that becomes outdated the moment a payer changes a guideline without notice (PM360, 2018). Coverage for treatments administered via infusion or injection, or across different sites of care, adds another layer of complexity that electronic databases simply cannot resolve without human verification (Infinitus, 2025).
When a benefit verification returns incomplete or inaccurate results, someone still has to call the payer to reconcile the information. The automation did not eliminate the call; it just delayed it. For a patient whose prior authorization is pending and whose case hits an unresolved verification gap late on a Friday afternoon, that delay is not a process footnote — it is days before the next opportunity to advance their therapy access.
Automating a flawed or incomplete process does not fix the process. At best, it accelerates the dysfunction. At worst, it pushes the patient further from therapy while giving the appearance of progress.
What Technology Should Actually Be Doing
None of this is an argument against innovation. The right use of automation in patient support programs has transformative potential, and the healthcare industry genuinely needs it. Administrative burden is one of the most significant contributors to provider burnout, and AI tools that handle documentation, status updates, scheduling, and routine eligibility checks free clinical staff to focus on the interactions that require judgment, empathy, and problem-solving. Hospitals that have implemented AI-assisted workflows report a return on investment of $3.20 for every $1 spent, often within 14 months of implementation (Litslink, 2026).
The question is not whether to deploy technology. The question is where. Technology earns its place in the patient journey when it handles what patients actually find burdensome: waiting days for a benefit verification response, not knowing where their case stands in the prior authorization process, sitting on hold to get a status update on something that could be communicated automatically. Those are solvable problems. Routing logic that identifies complex cases and sends them immediately to a credentialed specialist, rather than queuing them behind a fully automated process that cannot handle exceptions, is a meaningful advancement. Real-time dashboard visibility that surfaces bottlenecks in the patient access funnel before they compound into drop-off is a meaningful advancement.
Deploying a chatbot to manage patient questions that require judgment, clinical context, or payer-specific knowledge creates friction where there should be clarity. Committing to 100% electronic benefit verification on specialty therapies where payer data is demonstrably incomplete creates the appearance of efficiency while the actual resolution happens off-platform, in a phone call no one is tracking.
Building on a Foundation That Works
The future of patient support belongs to programs that combine smart automation with verified human intelligence. Technology should front-load intake, accelerate timelines on well-defined processes, flag exceptions immediately, and route complex cases to the people best equipped to resolve them. Human expertise should be applied where it matters most: benefit investigations that require payer outreach, prior authorization appeals with meaningful clinical documentation, and the direct patient communication that determines whether someone stays engaged with their therapy or quietly falls through the gap.
That combination — automation deployed thoughtfully, human support applied strategically — is what produces outcomes that can actually be measured. A 99% benefit verification accuracy rate. A four-hour turnaround on the cases that clear cleanly. A 90% overturn rate on prior authorization denials through credentialed appeals support. These are not outcomes that come from speed alone. They come from a delivery model where the technology and the people behind it are both accountable for the same result.
The pressure to automate is not going away. But the programs that will define the next generation of patient access are not the ones that deployed the most automation. They are the ones that deployed it in the right places.
References
EVERSANA. (n.d.). ACTICS eAccess: Electronics benefits verification. https://www.eversana.com/products/actics/actics-eaccess/
Infinitus. (2025, January 29). Benefit verification automation. https://www.infinitus.ai/solutions/benefit-verification/
Litslink. (2026, January 22). AI in healthcare statistics: Key trends shaping 2025. https://litslink.com/blog/ai-in-healthcare-statistics-and-trends
PCG Software. (2025, December 23). 10 AI chatbot use cases for payer organizations in 2026. https://www.pcgsoftware.com/ai-chatbots-in-healthcare
PM360. (2018, April 20). Using artificial intelligence to expedite the electronic benefit verification process. https://www.pm360online.com/using-artificial-intelligence-to-expedite-the-electronic-benefit-verification-process/
Shah, P., et al. (2025). Patient-facing chatbots: Enhancing healthcare accessibility while navigating digital literacy challenges and isolation risks — a mixed-methods study. SAGE Open Medicine. https://pmc.ncbi.nlm.nih.gov/articles/PMC12041682/
TempDev. (2025, May 28). 65 key AI in healthcare statistics. https://www.tempdev.com/blog/2025/05/28/65-key-ai-in-healthcare-statistics/
