The Plot Thickens: AI Agent Callers in Healthcare

From September 9th to October 6th 2026, weekly calls from Meta's Muse AI agent grew eightfold in Pindrop's data. 

Among the industries we analyzed, healthcare received the second-highest volume of calls, behind only financial institutions. Of the AI agent calls we analyzed, 94% were labeled as “good” agents, including healthcare providers' own agents. 

Providers send AI agents to payers to check benefit eligibility, claim status, and prior authorizations. Payers, meanwhile, must protect patient information while still managing legitimate requests. 

Without the right tools, a payer may not be able to tell a provider’s legitimate agent from a malicious bot. That creates a dilemma: If payers refuse every agent, providers have to make the same calls manually. If provider agents go unverified, payers risk exposing sensitive data to bad actors.

As agentic technology advances, tools and workflows that interact with it must keep up. Here’s how we’d work through it. 

First, determine if the caller is a human or agent

The first thing an organization must do is determine whether the caller is human, but that’s harder with every passing day. Voice agents now work through phone menus, press keypad tones, and speak convincingly with live staff. 

Among the Muse calls Pindrop analyzed, 47% pressed keypad digits to navigate the IVR, and 64% reached a human agent, spending a median of 2 minutes 18 seconds in conversation. The person on the other end may never realize the caller is automated, and some companies discourage asking, or even bar customer service staff from asking callers whether they are AI.

Multiply this situation by thousands of calls and then consider the even more difficult question organizations have to ask: If we can detect a caller as an agent, what do we do with it?

To block, or not to block?

Some organizations say proceed with caution, especially for low-risk interactions.¹ On the surface, many provider agents seemingly fall into this low-risk category (such as routine checks on claim status or eligibility or prior authorization requests), especially when compared to tasks that touch credit card information or authorize payments. 

Health insurers are already seeing agents in action. Steve Rullo, chief digital and technology officer at Guardian Life, said at InsureTech Connect in October that providers have started using AI to verify patient benefits, sometimes working through whole lists of patients in a single call. Rather than reviewing every transaction, Guardian keeps a human in the loop while moving toward governance that flags exceptions.

Low-risk calls can still be reconnaissance

Malicious bot callers often appear benign because there is no immediate fallout. Bots routinely conduct reconnaissance, mapping out IVRs, discovering authentication gates, and collecting patient information. Data connected to claims status or prior authorizations often contain Protected Health Information (PHI) or Personally Identifiable Information (PII), which immediately raises the stakes. 

Bad actors use bots to compile sensitive data for larger downstream attacks that lead to even greater PHI breaches and financial loss.

Refusing every agent is “safe”

For many organizations, the safest path is to block agents altogether until they figure out how to accurately verify who the agent works for, and what it is allowed to do. 

Financial institutions are already making their stance official. JPMorgan Chase CEO Jamie Dimon said the bank wouldn’t let just any third-party agent into its systems, and Charles Schwab's CEO said the firm is blocking unauthorized agents. As of this article’s publication date, I did not find official statements from health plan leaders regarding accommodations or bans for agents.

While blocking agent callers is the safest approach, it comes at a cost, primarily to the individuals who send them. In provider-to-payer engagements, for example, prior authorization requests are a major source of administrative burden to providers. In the American Medical Association's 2025 survey of 1,000 physicians, practices completed an average of 40 prior authorizations per physician per week, spent 13 hours of physician and staff time on them, and 40% employed staff dedicated only to prior authorization, despite efforts and pledges to reform related policies and requirements.

If payers declare themselves bot-free zones, agents may not be able to alleviate this burden for provider staff. The result: Providers are right back where they were before, making manual phone calls. 

Let in the agents you can verify

In order to accommodate agentic technology, healthcare organizations need to validate three things about every agent caller:

• Authenticity. Is this agent what it claims to be? Pindrop registered the voices of Muse's two agents in an AI Voice Consortium of more than 5,000 AI voices, and every Muse call in the analysis matched one of them.² That identifies a specific agent, not just the presence of a bot.
• Identity. Which organization and person is the agent working for? A payer needs to know that a call is from a real provider's office and that it is authorized before it shares coverage or claim details.
• Intent. What is the agent asking for, and does that fall within what it is authorized to do? Checking eligibility or claim status is a different request from changing member information.

With those answers, organizations can handle good agents and bad agents differently. Good agents are serviced, perhaps even by the organization's own AI agents, while bad agents are stopped. A request that goes past what the agent is authorized to do triggers a step-up, such as confirming the person behind it.

Where Pindrop BotStopper^TM technology fits today

Pindrop is at the forefront of this challenge. Pindrop BotStopper™ technology detects AI agents and automated callers early in the call, down to specific agents like Muse's, so organizations can start their AI agent handling journey today. 

There’s much more to come as we help organizations verify, authorize, and serve the good agents while stopping the bad ones. 

Learn about how Pindrop helps healthcare orgs defend against AI-powered attacks.

‍