Agentic AI Is Transforming Healthcare - But It Still Needs a Human Touch

We recently spent hours trying to resolve a complex email analytics issue with HubSpot – only to find ourselves trapped in a chatbot loop with no clear path to a human agent who could actually help. That experience, combined with Rajil Vohra's recent guest post on agentic AI in healthcare billing, led us to go deeper on where AI-driven customer service stands today in healthcare – and where it falls short.

Agentic AI – artificial intelligence that can autonomously observe, decide, and act across multi-step workflows – is reshaping healthcare customer service. Unlike earlier chatbots, agentic systems navigate complex processes end to end, from verifying eligibility to resolving billing inquiries in real time. AI-powered health tech startups captured 54% of all digital health funding in 2025, and Deloitte reports 85% of healthcare executives plan to increase agentic AI spending over the next three years.

The major players are moving fast. Cedar launched Kora, a voice-based AI agent for patient billing projected to automate 30% of inbound calls. Hyro raised $45 million to scale conversational AI across health systems. Commure secured $200 million for its AI-powered revenue cycle platform. Salesforce rolled out Agentforce Health, and Waystar is building an end-to-end agentic revenue cycle network.

The use cases gaining traction center on healthcare payments and revenue cycle management – where providers collectively spend over $140 billion annually. AI agents handle scheduling, answer copay and deductible questions, identify payment options, route patients to financial assistance, and process billing inquiries around the clock. McKinsey estimates agentic AI could reduce cost-to-collect by 30 to 60 percent.

For routine interactions, agentic AI excels. It eliminates hold times, operates 24/7, supports multiple languages, and delivers consistent answers to straightforward questions. Cedar's Kora can detect sentiment, explain charges conversationally, and escalate to a live agent when needed – all while maintaining HIPAA compliance.

But healthcare billing is rarely straightforward. A patient whose surgery is denied 48 hours before the procedure due to a coding error, surprise balance bills involving multiple providers, disputes over medical necessity, coordination of benefits across overlapping coverage – these situations demand contextual judgment that today's AI doesn't possess. Vohra's article last issue makes clear that agentic systems should escalate to humans when judgment or intervention is required. But as our recent HubSpot experience illustrates, organizations must make it obvious how to reach a human and for whatsituations, or customers with complex problems will burn hours fighting through automated responses.

There's also a fundamentally human dimension. ECRI flagged AI chatbot misuse as the top health technology hazard of 2026, and 60% of Americans remain uncomfortable with AI playing too large a role in their care. Patients facing a confusing medical bill don't just need correct information – they need compassion and the reassurance that someone is listening.

The most effective healthcare organizations won't replace human support with AI; they'll use it to handle the routine so real people can show up where it matters most.