UnitedHealthcare has expanded the availability of its generative AI assistant, Avery, to approximately 21 million members across its commercial, Medicare, and Medicaid health plans. The expansion follows an initial rollout that reportedly allowed the assistant to resolve 90% of member interactions without requiring assistance from a human representative.
Introduced in March, Avery was initially available to approximately 6.5 million employer-sponsored plan members and 160,000 Medicare Advantage members. The insurer had originally targeted 20.5 million eligible members by the end of the year, a milestone it has now exceeded.
The expanded deployment reflects UnitedHealthcare’s efforts to use generative AI to improve customer service, simplify access to insurance information, and reduce the need for members to contact human representatives.
Avery Provides Personalised Insurance Information
Available through the UnitedHealthcare mobile application and myuhc.com, Avery allows eligible members to ask questions about their health insurance coverage and receive responses based on their individual plans.
The assistant can retrieve information from the insurer’s benefits, provider, and claims systems to answer questions about deductibles, out-of-pocket expenses, provider networks, and medical claims.
Members can also review plan balances, check available benefits, estimate the costs of medical services, and determine whether healthcare providers are included in their insurance networks. Provider information can include current network status, contact details, and locations.
UnitedHealthcare developed Avery by analysing patterns in customer service conversations and telephone interactions. The company used this information to fine-tune its AI systems to respond to common member inquiries using existing digital data.
The assistant also supports employers and insurance brokers by answering questions about insurance products, comparing plans, and processing uploaded enrollment and eligibility documents.
AI Integration Extends to Human Customer Service
Beyond handling routine inquiries, Avery is designed to support UnitedHealthcare’s existing customer service operations.
When a member requires additional assistance, the assistant can transfer the conversation to a human advocate while providing a summary of the previous interaction. Representatives can then access relevant member information and conversation history without requiring customers to repeat details or resubmit documents.
This approach is intended to make the transition between automated and human support more efficient while reducing friction for members with more complicated requests.
UnitedHealthcare is also developing additional capabilities that could allow Avery to contact healthcare providers and arrange appointments on behalf of members.
Although appointment scheduling was included among the assistant’s initially announced capabilities, the insurer has not clarified how widely direct provider calling is currently available.
Reported 90% Resolution Rate
UnitedHealthcare has reported that Avery resolves approximately 90% of member interactions without requiring human assistance. The remaining 10% are transferred to human advocates.
However, the insurer has not disclosed the methodology used to calculate this resolution rate or provided detailed figures on the number of conversations handled.
The reported 21 million members represent those eligible to access Avery, rather than confirmed active users. UnitedHealthcare has not published figures showing how many members have actually used the assistant or how frequently they interact with it.
The company evaluates Avery using several performance indicators, including accuracy, completeness, resolution rates, and response times. Separate numerical results for these measures have not been made publicly available.
Responsible AI and Healthcare Oversight
UnitedHealthcare acknowledges that Avery can sometimes provide incomplete or inaccurate responses. Its documentation makes clear that the assistant provides general insurance information rather than medical advice, formal benefit determinations, guarantees of coverage, or healthcare service authorisations.
Actual coverage remains subject to the terms, conditions, limitations, and exclusions of each member’s insurance plan. Members are also advised that Avery should not replace guidance from qualified healthcare professionals.
The assistant operates under UnitedHealth Group’s broader Responsible AI governance framework, which addresses model performance, data usage, fairness, privacy, and potential bias.
According to UnitedHealth Group’s 2025 proxy statement, its Artificial Intelligence Review Board includes specialists in technology, clinical care, medical ethics, data science, cybersecurity, privacy, legal matters, and member advocacy.
The board can approve AI models for deployment or require their temporary or permanent removal from production. The company’s Responsible AI Program Office also maintains an inventory of AI models and oversees the required review processes.
UnitedHealth Group’s AI principles emphasise reliability testing, continuous monitoring, bias assessments, and safeguards for sensitive information.
Privacy and Transparency Remain Important Considerations
Despite its broader AI governance framework, questions remain about how Avery handles sensitive member information.
UnitedHealth Group states that its AI applications must comply with applicable privacy and security requirements. However, publicly available documentation has not detailed how long Avery conversations are retained, whether member interactions are used for additional model training, or whether independent assessments of the assistant’s accuracy have been conducted.
Greater transparency around these practices could help members better understand how their personal information is processed and how reliably the assistant handles insurance-related questions.
The Growing Role of AI in Healthcare Administration
UnitedHealthcare’s expansion of Avery demonstrates how generative AI is becoming increasingly integrated into healthcare administration and customer support.
By combining personalised insurance information, automated customer service, and human representative escalation, the insurer is attempting to make routine interactions more efficient while maintaining access to human assistance when necessary.
However, the reported resolution rate alone does not establish whether members consistently receive accurate answers or achieve better outcomes. As AI assistants take on more responsibilities, insurers will need to demonstrate that automation delivers reliable service while maintaining appropriate privacy protections and human oversight.
Source: https://www.artificialintelligence-news.com/news/unitedhealthcare-ai-assistant-avery/
