AI Agents for Healthcare Teams: Scheduling, Billing, and Follow-Up Care

AI Agents for Healthcare Teams: How Scheduling, Billing, and Follow-Up Care Are Getting Automated

Healthcare has spent two years talking about generative AI. That conversation just crossed a real threshold: half of U.S. healthcare organizations now say they've actually implemented gen AI, according to McKinsey's latest survey of US healthcare leaders, conducted between September 17 and October 17, 2025 across 50 payers, 50 clinical-care organizations, and 50 health services and technology (HST) firms. More than 80 percent have already pushed their first use cases out to end users.

The next wave isn't about generating text — it's about agents that actually do the administrative work healthcare has never had enough staff to keep up with: scheduling, billing, and the follow-up calls that fall through the cracks. Here's where that's already paying off, straight from the primary research.

Why Billing and Scheduling Are the Obvious Starting Point

Ask healthcare leaders where AI has the most potential, and administrative efficiency tops the list — ahead of clinical productivity, patient engagement, and software infrastructure, according to McKinsey's survey. That tracks with where the money and the pain actually are. Revenue cycle management (RCM) — everything from patient scheduling through claims and collections — typically costs an at-scale health system 3 to 4 percent of its revenue, and McKinsey's analysis of the sector puts collective US health system spending on RCM north of $140 billion a year.

The waste inside that number is stark: nearly 20 percent of claims are denied on average, and as many as 60 percent of denials are never even appealed — money health systems have simply written off. McKinsey estimates that enabling the revenue cycle with agentic AI could cut the cost to collect by 30 to 60 percent. For a health system billing $6 billion a year, shaving even one to two percentage points off a typical 3.5–4.0 percent cost-to-collect rate works out to $60 million to $120 million in savings.

Where Healthcare Teams Are Actually Deploying Agents

Billing and Collections: Starting at the Back End, Not the Front

Most health systems aren't handing agentic AI the whole revenue cycle at once — and McKinsey's research suggests that's the right call. The back end of RCM (accounts-receivable follow-up, underpayment management, denials management, cash posting) is labor-intensive, rules-governed work where staffing is the main bottleneck, which makes it a natural fit for autonomous agents while human staff manage the exceptions. It's also lower-risk: back-end billing work doesn't touch a live patient interaction the way scheduling or clinical documentation does, so it's a safer place to test and refine an agent before expanding its scope.

Agentic AI Adoption Is Still Early, But Moving Fast

Only 19 percent of healthcare organizations surveyed have reached agentic AI implementation, per McKinsey — but 51 percent report they're actively running agentic AI proofs of concept, and just 1 percent have no plans to pursue AI agents at all. Adoption isn't even across the industry: HST firms are furthest along, while payer organizations lag, with implementation below 50 percent even for standard gen AI.

Follow-Up Care and Patient Engagement

Patient and member engagement ranks as one of the highest-potential domains for gen AI in McKinsey's survey, even though it's currently one of the least-implemented — a gap the report calls out directly as where organizations should focus next. That's the follow-up work agents are increasingly built for: automated appointment reminders, post-visit check-ins, and closing the loop on care instructions that used to depend on a staff member remembering to make a call.

What's Actually Slowing Adoption Down

It isn't hesitation about the technology's usefulness — 82 percent of surveyed healthcare leaders expect a positive return on their gen AI investment, the highest share McKinsey has recorded since it started asking. The real barriers are operational: integration challenges into legacy healthcare IT systems and a lack of internal AI capabilities rank as the top two roadblocks to scaling, ahead of the risk and safety concerns that used to dominate the conversation. In other words, most healthcare organizations now believe in the payoff — they're stuck on plumbing it into systems that were never designed for it.

What This Means for Healthcare Teams Going Forward

The organizations getting real traction aren't trying to automate clinical judgment. They're pointing agents at the back-end, rules-based grind — claims, denials, collections, scheduling logistics, routine follow-up — where the volume is high, the patterns are learnable, and a human is there to handle whatever doesn't fit the pattern. That's a very different bet than a full clinical-AI overhaul, and it's exactly the sequencing McKinsey's own research recommends: prove value in a contained, lower-risk area, then expand.

Frequently Asked Questions

Do AI agents replace billing and scheduling staff in healthcare? Not in the deployments McKinsey documents. The pattern is agents absorbing high-volume, rules-based tasks like claims follow-up and cash posting, with staff refocused on exception handling and more complex, higher-value work rather than being replaced outright.

What's actually blocking agentic AI adoption in healthcare right now? Integration into existing healthcare IT systems and a shortage of internal AI capability — not doubts about ROI. 82 percent of healthcare leaders in McKinsey's Q4 2025 survey expect a positive return, the highest figure the survey has recorded.

Where should a healthcare organization start with agentic AI? The back-end revenue cycle — accounts receivable follow-up, denials management, underpayment recovery — is the highest-leverage, lowest-risk starting point, according to McKinsey's analysis, because it's rules-governed, high-volume, and doesn't touch live patient interactions.