Two Years In: What a Mature AI Voice-Agent Deployment Actually Looks Like
By Genevieve Carrenard
WellSpan Health’s AI agent now handles more than 160,000 patient calls a month and is expanding from scheduling into post-discharge follow-up. Here’s what that trajectory can tell your call center about evaluating AI investment.
Most of what gets published about AI in the medical call center is still speculative: a vendor’s roadmap with no production data behind it. WellSpan Health’s deployment isn’t speculative. Ana, the generative AI voice agent the Pennsylvania health system built with Hippocratic AI, has been live for nearly two years. It now handles more than 160,000 patient calls and 7,000-plus hours of conversation every month. On July 30, WellSpan and Hippocratic AI announced they’re expanding the partnership platform-wide, which is a useful moment to ask what a mature AI deployment actually looks like once the pilot phase is over.
From Scheduling Into Follow-Up
Ana started, like most call center AI, on the easiest ground: inbound calls and primary-care appointment scheduling. The expansion moves it into harder territory: outbound post-discharge follow-up and chronic-disease check-ins, the kind of proactive outreach that used to depend on staff finding time between inbound calls. WellSpan is also trialing voice AI tools to support clinicians with triage tasks and launching what Hippocratic AI calls an “Orchestrator,” meant to help care teams navigate a patient’s care plan and history rather than just book their next visit. None of this happened overnight. Hippocratic AI has a dedicated team embedded on WellSpan’s York, Pennsylvania campus, co-developing the next generation of agents instead of shipping a static product and walking away.
What This Means for Your Call Center
You don’t need WellSpan’s budget to learn from WellSpan’s arc. The pattern here is a reasonable sequence for any medical call center sizing up an AI vendor, not just a large health system: start with scheduling, prove it, then expand into outbound and higher-stakes work only after the easy ground is solid. Before you sign anything, ask a vendor to show you what happened at month three, month 12, and month 24 of someone else’s deployment, not just the launch numbers. Ask what the escalation path looks like when the AI is wrong. Ask who’s actually watching outcomes. Answering the phone and doing the job well are two different things. Two years into a real deployment, that’s still the distinction that matters most.
Source: Fierce Healthcare, “WellSpan Health, Hippocratic AI ink ‘multi-year’ partnership with plans to co-develop clinical AI agents,” July 30, 2026, corroborated by HIT Consultant and WellSpan’s own newsroom.
News
Twenty-eight percent of patients say they’ve put off canceling or rescheduling an appointment because doing so required a phone call, according to a new Tebra survey — a reminder that phone friction cuts both ways for call volume.
Read more: The intake
A ransomware group calling itself The Gentlemen hijacked AnMed Health’s Facebook page on Aug. 11 to claim it stole 6TB of data, including highly sensitive patient records; the South Carolina hospital system says the claims are unverified and under investigation.
Read more: HIPAA Journal
HHS has pushed final action on its proposed HIPAA Security Rule overhaul — mandatory encryption, multi-factor authentication, and 72-hour breach reporting — to at least July 2027, giving covered entities and business associates, including medical answering services, a longer runway to prepare.
Read more: Fierce Healthcare
Send us your healthcare call center articles and news for the next issue of Medical Call Center News.
A Thought for Today
“The most important practical lesson that can be given to nurses is to teach them what to observe — how to observe.”
— Florence Nightingale, Notes on Nursing (1859)