Should Patients Know When AI Is Calling About Their Bill?
Hospitals are debating whether to tell patients when AI is involved in their clinical care. That debate has not yet reached the revenue cycle, where automated and AI-driven calls about appointments, estimates, and balances are scaling quickly. Revenue cycle leaders should set a disclosure policy for patient financial communications now. Federal rules already govern much of how these calls are made, and patient trust in billing communications is harder to rebuild than it is to protect.
Key takeaways
Epic's Hello World platform helped health systems prevent more than 3.4 million no-shows over the past 12 months, and Epic has added Branded Calling, which shows the health system's name, logo, and reason for the call before a patient answers.
Health system leaders publicly disagree on whether patients should be told when AI is involved in their care. Rhode Island became the first state to require providers to notify patients when they use an AI scribe.
The FCC ruled in February 2024 that AI-generated voices count as "artificial" voices under the Telephone Consumer Protection Act (TCPA), which generally requires prior express consent for those calls.
The TCPA's healthcare exemption for certain automated calls excludes billing, accounting, and debt collection content, so patient financial calls generally need consent.
Branded caller ID tells patients who is calling. It does not tell them what is calling, and that is the gap a disclosure policy should close.
Are hospitals using AI to call patients about their bills?
Yes, and the volume is growing fast. Epic's Hello World platform sends automated texts and voice calls that let patients confirm, get reminders for, or reschedule appointments without staff placing the calls, using each patient's preferred channel from the medical record (Becker's). Epic reports the platform helped prevent more than 3.4 million no-shows nationwide over the past year.
The new Branded Calling feature, announced September 23, displays the health system's name, logo, and reason for the call on the patient's phone before pickup. It works across all major U.S. carriers through an integration with Vonage. Community Health Network in Indianapolis reported a 54% improvement in voice reminder answer rates and a 6% improvement in its overall no-show rate. Ochsner Health reported that SMS rescheduling saved call center staff more than 1,000 hours over 10 months.
AI voice agents are scaling alongside that automation. This summer, Becker's reported that WellSpan's voice agent handles more than 160,000 patient calls a month, MUSC has passed 2.2 million calls since 2024, and Tampa General cut call abandonment from 34% to 14.9%.
Not every automated call uses AI. Many reminder calls are rules-based and prerecorded. From the patient's side of the phone, the distinction is often invisible, and that is part of what makes disclosure a policy question and not just a technical one.
Do hospitals have to tell patients when AI is involved?
There is no single national rule, and health system leaders disagree on what the standard should be. At Mass General Brigham's World Medical Innovation Forum on September 23, a panel split on the question (Becker's).
Mass General Brigham's chief health information officer pointed out that hospitals already use AI without telling patients. The top of a standard EKG readout, for example, is generated by an algorithm. She also noted that ambient documentation consent is usually consent to record the visit, not consent to use an AI model. Her recommendation was to start with the purpose of disclosure: preserving the patient's right to refuse, supporting a second opinion, or demonstrating transparency.
A startup CEO on the same panel took the opposite view. He argued that patients should be told when they are having a conversation with a machine.
State law is starting to fill in. Rhode Island became the first state to require providers to notify patients when they use an AI scribe. Twelve states, including Massachusetts, require two-party consent to record audio. None of this debate has focused on billing calls yet, which is why revenue cycle leaders should not wait.
What federal rules already apply to automated and AI voice calls?
The Telephone Consumer Protection Act already covers much of this territory. In February 2024, the FCC issued a declaratory ruling that AI-generated voices are "artificial" under the TCPA, which means businesses using them generally need prior express consent from the person they are calling (Wilson Sonsini). The FCC also made clear that the law has no carve-out for technology designed to act like a live agent (HWG).
Healthcare providers have a narrow exemption for certain automated calls and texts to mobile phones, created by the FCC in 2015. It covers messages with a healthcare treatment purpose, such as appointment reminders, pre-registration instructions, lab results, and post-discharge follow-up. The exemption excludes calls that include accounting, billing, debt collection, or other financial content, and it requires the caller to state the provider's name and contact information at the beginning of a voice call (Epstein Becker Green).
The practical implication for revenue cycle leaders is that an automated or AI voice call about a balance, estimate, or payment plan generally falls outside the healthcare exemption and needs documented consent. I am not a lawyer, and this is not legal advice. Your compliance and legal teams should review how consent is captured and documented for each type of automated patient financial call.
Why does disclosure matter more for billing calls than clinical ones?
A clinical AI tool usually operates inside an encounter the patient chose to have. A billing call arrives unannounced, asks the patient to act, and often involves money. That is the same shape as the scams patients are warned about every week.
In RCM 2030, I wrote that your patient is also your customer, and that patients benchmark every interaction against banking, retail, and travel. Banks and retailers have trained consumers to expect to know who they are dealing with before they share financial information. A patient who discovers partway through a call that they were talking to software may reasonably question whether the call was legitimate at all. That patient is less likely to answer the next call and less likely to pay through the channel you invested in.
Disclosure also protects the human side of the revenue cycle. I wrote in the book that if AI is the engine, financial counselors are the pilots. Patients still want a real person when the stakes are high, the options are confusing, or trust is fragile. An AI agent that identifies itself and offers a transfer to a person keeps that path open. One that doesn't can leave a patient negotiating a payment plan without knowing a counselor was available.
What does branded caller ID change?
Branded Calling solves a real problem. Patients ignore unknown numbers, and Community Health Network's 54% improvement in answer rates shows how much a verified name on the screen matters.
It solves only part of the problem, though. Branded caller ID tells the patient who is calling. It does not tell them what is calling, whether that is a person, a prerecorded message, or an AI agent that can respond to questions. As answer rates rise, more patients will reach automated systems, which makes the disclosure question more pressing.
How should revenue cycle leaders set an AI disclosure policy for patient financial communications?
Start with purpose. Following Mass General Brigham's recommendation, decide what disclosure is meant to accomplish for each call type. For a billing call, the purposes are usually trust, the patient's ability to reach a person, and compliance with consent rules.
Disclose at the start of the call. Identify the health system and the fact that the patient is speaking with an automated assistant in the first few seconds, before any account or balance information is discussed. The healthcare exemption already requires provider identification at the beginning of exempt calls; extending the same practice to financial calls is a reasonable standard.
Always offer a person. Every automated financial call should include a clear, fast way to reach a financial counselor or billing representative. Track how often patients use it. A high transfer rate on one call type is useful information about where automation isn't serving patients.
Tie consent records to call types. Know which patients have consented to automated calls about financial matters, how that consent was captured, and when. Registration and portal enrollment workflows are the most common places to capture it, and both are revenue cycle workflows.
Use the same language everywhere. Publish one plain-language statement about how your organization uses automated calls and what it will never ask for by phone. Use identical wording on statements, in the patient portal, and in call scripts so patients can recognize legitimate outreach.
Review vendor contracts. Confirm whether your communications or voice AI vendor supports disclosure scripts, opt-out handling, and consent documentation, and who carries the risk if a call type falls outside what the patient consented to. My free AI Renewal Checklist includes the questions to ask before your next renewal.
What will patient billing communications look like by 2030?
By 2030, I expect most routine patient financial communication to be automated: appointment and payment reminders, estimate delivery, payment plan setup, and simple balance questions. I also expect disclosure to be standard practice, driven by some combination of state law, payer and accreditation expectations, and patient demand.
The organizations that set a disclosure policy now will be able to automate faster, because patients will trust the channel. In RCM 2030, I described the goal as communication tailored to each patient, where some need a payment plan link in a text reminder and others need a phone call, and where empathy and revenue align (p. 65). Telling patients when they are talking to a machine is part of that alignment.
The full framework for patient financial engagement, including segmentation and the future role of financial counselors, is in RCM 2030: Strategy and Survival for Revenue Cycle Leaders.
Frequently asked questions
Do hospitals have to tell patients when AI is used? There is no single national requirement. Rhode Island requires providers to notify patients when they use an AI scribe, and 12 states require two-party consent to record audio. Federal TCPA rules generally require prior express consent for calls that use an AI-generated or prerecorded voice. Health systems should consult their legal teams about the rules in their states.
Are AI voice calls legal under the TCPA? AI voice calls are not banned, but the FCC ruled in February 2024 that AI-generated voices are "artificial" under the TCPA. Calls using them generally require the recipient's prior express consent unless an exemption applies.
Does the TCPA healthcare exemption cover billing calls? No. The FCC's 2015 healthcare exemption covers certain treatment-related calls, such as appointment reminders and lab results, but excludes calls with accounting, billing, debt collection, or other financial content.
What is Epic Branded Calling? Branded Calling is a feature of Epic's Hello World communication platform that displays a health system's name, logo, and reason for the call on a patient's phone before they answer. Epic announced it on September 23, 2026, through an integration with Vonage.
How many patient calls are AI voice agents handling? Volumes vary by health system. WellSpan's voice agent handles more than 160,000 patient calls a month, and MUSC has passed 2.2 million calls since 2024, according to Becker's reporting in 2026.
What should an AI disclosure statement on a billing call include? At minimum, the name of the health system, a statement that the patient is speaking with an automated assistant, and a clear way to reach a person. It should come at the start of the call, before any account or balance information is discussed.
Why does AI disclosure matter for hospital revenue? Patients who don't trust billing communications are less likely to answer calls, use portals, or pay through automated channels. Clear disclosure protects the answer rates and payment behavior that automation is meant to improve.
