UnitedHealthcare has expanded access to Avery, its generative AI-powered health insurance assistant, to 21 mn members across commercial, Medicare and Medicaid plans, extending the technology beyond its initial employer-sponsored and Medicare Advantage rollout.
The expansion follows Avery’s introduction in March 2026, when approximately 6.5 mn employer-sponsored plan members and 160,000 Medicare Advantage members were eligible to use the assistant. UnitedHealthcare initially targeted 20.5 mn eligible members by year-end.
- UnitedHealthcare has expanded Avery to 21 mn eligible members, up from approximately 6.7 mn at its March launch, exceeding its original year-end target of 20.5 mn.
- The generative AI assistant reportedly resolves 90% of member interactions without human intervention, although the insurer has not disclosed usage figures or its calculation methodology.
- Avery provides personalized benefits, claims and cost information, while UnitedHealthcare is developing additional agentic AI functions and applying its Responsible AI governance framework.
Craig Kurtzweil, UnitedHealthcare’s chief data and analytics officer, subsequently said Avery had resolved 90% of member interactions without requiring assistance from a human representative.
The remaining 10% were transferred to human advocates. UnitedHealthcare has not disclosed how it calculates the resolution rate or provided independent performance data.
The 21 mn figure represents members eligible to access Avery, rather than actual users. The insurer has not published the number of members who have interacted with the assistant or the total volume of conversations handled.
How UnitedHealthcare’s Avery AI assistant works
Avery is available through the UnitedHealthcare mobile application and myuhc.com, allowing eligible members to ask questions about their health coverage and receive information specific to their insurance plans.
The assistant retrieves information from UnitedHealthcare’s benefits, provider and claims systems to answer questions about deductibles, out-of-pocket expenses, medical costs, claims and available benefits.
Members can check their plan balances, obtain estimates for medical services and determine whether particular healthcare providers participate in their insurance network.
Provider searches can include current network status, contact information and location, helping members verify their options before obtaining treatment.
UnitedHealthcare developed Avery by analyzing patterns in customer service conversations and telephone inquiries. The insurer previously said its AI systems had been fine-tuned to respond to common member questions using information from its digital platforms.
The assistant also supports brokers and employers with insurance product inquiries, plan comparisons and the processing of uploaded enrollment and eligibility documentation.
Customer service representatives can use Avery to retrieve information from multiple internal systems and document member interactions.
When an inquiry requires human assistance, the AI assistant transfers the member to a customer service advocate and provides a summary of the conversation.
Representatives can review relevant member information and previous interactions, reducing the need for customers to repeat questions or submit the same documents again.
UnitedHealthcare develops AI agents for appointment scheduling
UnitedHealthcare is working on additional capabilities that would allow Avery to contact healthcare providers and arrange medical appointments on behalf of members.
Kurtzweil described the planned functionality in an interview with Fierce Healthcare, indicating that the insurer intends to extend Avery beyond answering questions into performing administrative tasks.
The company’s March announcement had already identified appointment scheduling among Avery’s capabilities. However, UnitedHealthcare has not clarified whether direct provider calling is currently available to all eligible members or remains under development.
The distinction is relevant because automated scheduling requires an AI system to interact with external healthcare organizations, rather than simply retrieve information from existing insurance records.
For members, such functionality could reduce the administrative work involved in identifying participating providers and arranging appointments. Its eventual availability and operating limits have not been specified.
AI accuracy, insurance coverage and member privacy
UnitedHealthcare evaluates Avery using several performance measures, including response accuracy, completeness, resolution rates and response time.
Although the company has reported a 90% interaction resolution rate, it has not disclosed numerical results for the other measures. The insurer acknowledges that AI-generated responses may contain errors or incomplete information.
Its official documentation states that Avery provides general information and does not constitute medical advice, a formal insurance benefit determination, a guarantee of coverage or authorization for healthcare services.
Insurance coverage remains governed by the terms, limitations, conditions and exclusions of each member’s health plan.
Avery also cannot replace advice from qualified healthcare professionals. These limitations apply even when the assistant retrieves information from UnitedHealthcare’s internal systems, as an AI-generated response is not itself a binding coverage decision.
UnitedHealth Group applies Responsible AI governance
UnitedHealthcare says Avery operates under UnitedHealth Group’s Responsible AI governance framework, which establishes requirements for model performance, data use, fairness, privacy and potential bias.
According to UnitedHealth Group’s proxy statement, its Artificial Intelligence Review Board includes technology specialists, clinicians, medical ethicists, data scientists, security experts, legal and privacy professionals, and member advocates.
The board can approve AI models for deployment or require their temporary or permanent removal from production.
UnitedHealth Group’s Responsible AI Program Office maintains an inventory of AI models and oversees their review under the company’s governance procedures. Its AI principles require reliability testing, continued performance monitoring, bias assessments and safeguards for sensitive information.
The company also states that AI applications must comply with applicable privacy and data security requirements.
UnitedHealthcare has not disclosed the results of Avery-specific bias assessments or independent accuracy evaluations in the publicly available materials reviewed.
Those materials also do not specify how long member conversations are retained or whether information collected through Avery is subsequently used for additional AI model training.
The insurer’s reported 90% resolution rate therefore provides an indication of how frequently interactions are completed without human assistance, but does not independently establish the accuracy of every response or the quality of the resulting member experience.
