UnitedHealth and Optum AI Strategy: How the Largest US Health Insurer Is Going Conversational in 2026

TL;DR#

UnitedHealth Group, the largest US health insurer, and its Optum health-services arm are building a conversational, AI-first member experience anchored by Avery, a generative-AI companion UnitedHealthcare launched in March 2026. Avery started with roughly 6.5 million employer-plan members and 160,000 Medicare Advantage members and is slated to reach about 20.5 million commercial, Medicare, and Medicaid members by year-end (UnitedHealth Group newsroom). The company says it is investing roughly $1.5–1.6 billion in AI in 2026, much of it inside OptumInsight, targeting near $1 billion in operating-cost reductions (Fierce Healthcare). For any organization studying ai healthcare intake, UnitedHealth's bet is instructive: the highest-value AI surface is not the back office but the first conversation a member has about coverage, benefits, and care. UnitedHealth also faces ongoing litigation over a separate, unrelated algorithm (nH Predict) used in post-acute coverage review, which the company disputes — a reminder that conversational intake and automated claims adjudication are different problems with very different stakes.

What Is UnitedHealth Group's AI Strategy in 2026?#

UnitedHealth Group's 2026 AI strategy is a roughly $1.5 billion enterprise-wide program to convert legacy systems into AI-driven platforms across UnitedHealthcare (insurance) and Optum (health services), with member-facing conversational tools and OptumInsight modernization as the headline investments. CEO Stephen Hemsley has framed the spend as transitioning older products toward "AI-first operations," with financial benefits expected to scale in the second half of 2026 (Healthcare Finance News).

The scale here is hard to overstate. UnitedHealthcare is the largest private insurer in the US and the biggest Medicare Advantage carrier, and UnitedHealth Group runs roughly 2,700 subsidiaries and affiliates (UnitedHealth Group on Wikipedia). Optum's 2024 revenues reached about $253 billion, and Optum Health served 4.7 million people under value-based care arrangements at the end of 2024. When a company this large standardizes on conversational AI for member contact, it sets a de facto template for what ai healthcare intake looks like across the industry — much as we have seen at integrated systems like Kaiser Permanente and rival payers like Aetna and CVS Health and Cigna.

How Does Avery Work as a Conversational Intake Layer?#

Avery works as a generative-AI companion that members talk to in natural language to handle benefits, coverage, and care-navigation tasks that historically routed through phone trees, FAQ pages, or human advocates. According to UnitedHealthcare, Avery can help members with coverage and personal benefits, appointment scheduling, cost estimates and plan balances, rewards and wellness programs, ID cards, over-the-counter benefits, provider search, claim approval status, and explanations of benefits (Fierce Healthcare).

Two details make Avery a useful case study for intake design specifically:

• Deflection with a human safety net. UnitedHealthcare reports that about 90% of the time, members who use Avery do not need an advocate, and Avery automatically transfers to a human when they do. That is the right architecture for a regulated, high-anxiety domain — the AI handles the routine "where do I start" questions and escalates the rest rather than trapping people in a bot.
• Phased, segment-aware rollout. Starting with roughly 6.5 million employer-plan members and 160,000 Medicare Advantage members before expanding to ~20.5 million across commercial, Medicare, and Medicaid lines shows the cohorts get sequenced by risk and complexity, not flipped on all at once.

This is the same pattern emerging across healthcare onboarding and intake — replacing static forms and IVR menus with a conversation that adapts. We have documented it in practice settings in our guides on replacing paper patient-intake forms with conversations and replacing clipboards with conversational forms.

Where Optum Fits: AI Beyond the Front Door#

Optum is the engine behind much of UnitedHealth's AI spend, with roughly a third of the 2026 investment directed at transforming OptumInsight and the rest funding AI across member services, claims processing, and administrative workflows (Becker's Payer Issues). Reported initiatives include a digital prior-authorization tool the company says reached a 96% approval rate in its early months and a new AI consulting unit, OptumAI, that has signed its first clients.

For an intake lens, the important distinction is between member-facing conversational AI (Avery, benefits navigation, care intake) and back-office automation (prior authorization, claims adjudication). The first is about helping a person express what they need and routing them correctly. The second is about operational throughput. They share infrastructure, but the design constraints and the regulatory exposure are very different — a point that becomes critical in the next section. Healthcare organizations evaluating conversational intake should be explicit about which problem they are solving; the public playbooks from telehealth-first players like Teladoc, Hims & Hers, Ro, and One Medical generally lead with the member-facing layer first.

The Sensitive Context: Claims-Denial AI Litigation#

UnitedHealth is also a defendant in ongoing litigation over a separate algorithm, and it is important to keep that case distinct from its conversational-intake work. A class action, Estate of Lokken v. UnitedHealth Group, alleges that UnitedHealthcare used nH Predict — a tool developed by Optum subsidiary naviHealth — to estimate post-acute care duration and end coverage based on those estimates rather than treating physicians' judgment (CBS News). In a March 2026 order, a federal magistrate judge in Minnesota directed UnitedHealth to turn over a broad set of documents, largely siding with plaintiffs on a motion to compel discovery (Becker's Payer Issues).

UnitedHealth disputes the characterization. The company has stated that nH Predict "is not used to make coverage decisions" and is instead "a guide to help us inform providers, families and other caregivers about what sort of assistance and care the patient may need" (CBS News). The litigation is unresolved, and the allegations are contested claims, not findings.

The takeaway for anyone designing ai healthcare intake is a governance one, not a verdict: conversational intake (helping a member describe their situation and find the right next step) and automated coverage adjudication (deciding what gets paid) are different functions with different risk profiles. Using AI to listen and route is a fundamentally lower-stakes use than using AI to deny. Keeping those layers separate — with humans firmly in the loop on coverage decisions — is the prudent design lesson the public record supports.

What This Means for Healthcare Intake Design#

The lesson from UnitedHealth and Optum is that conversational AI delivers its clearest, lowest-risk value at the front door — onboarding, benefits navigation, and care intake — where the job is to understand a member in their own words and route them correctly. Here is the practical pattern the public record suggests.

For practices, plans, and provider groups building their own intake, four principles follow:

1. Start at intake, not adjudication. Replace the form and the phone tree first. That is where conversational AI earns trust and deflects volume without high regulatory exposure.
2. Capture the "why," not just fields. A member who types "my mom was just discharged and I don't understand her new plan" needs follow-up questions, not a dropdown. This is exactly where conversational intelligent intake beats static forms.
3. Always provide a human path. Avery's 90% self-service rate works because the other 10% reaches a person. Design the handoff first.
4. Keep listening and deciding separate. Use AI to gather and route; keep humans accountable for coverage decisions.

This is the core of Perspective AI's view: AI-first member and patient experiences cannot start with a web form. A form flattens a worried caregiver into checkboxes; a conversation lets them explain the situation and surfaces what they actually need. Teams running these programs — see how we frame it for CX teams — use an AI interviewer agent to probe and an AI concierge agent to replace the form at the point of entry. You can browse member and patient intake studies or start a new conversational study to see the pattern applied. Peer examples across the sector — from Centene's Medicaid member experience to Cleveland Clinic's care navigation and GoodRx's prescription-savings flows — reinforce that the front door is where AI conversations pay off first.

Frequently Asked Questions#

What is Avery, UnitedHealthcare's AI companion?#

Avery is a generative-AI companion UnitedHealthcare launched in March 2026 to help members navigate coverage, benefits, scheduling, cost estimates, provider search, claim status, and explanations of benefits through natural-language conversation. It began with roughly 6.5 million employer-plan members and 160,000 Medicare Advantage members, with planned expansion to about 20.5 million members by year-end. When a member needs more help, Avery automatically transfers them to a human advocate.

How much is UnitedHealth Group investing in AI?#

UnitedHealth Group plans to invest roughly $1.5–1.6 billion in AI in 2026, according to company statements and reporting from Fierce Healthcare and Becker's. About a third is directed at transforming OptumInsight, with the remainder funding AI across member services, claims processing, and administrative workflows. The company has said it expects near $1 billion in operating-cost reductions in 2026, many of them AI-enabled.

Is UnitedHealth using AI to deny insurance claims?#

UnitedHealth faces a class-action lawsuit alleging its subsidiary's nH Predict algorithm was used to limit post-acute coverage, but those allegations are contested and unproven. UnitedHealth states that nH Predict "is not used to make coverage decisions" and serves as a guide for care needs. A federal court ordered broad document discovery in March 2026, and the case remains unresolved. This is separate from the company's member-facing conversational tools like Avery.

How is AI healthcare intake different from AI claims processing?#

AI healthcare intake uses conversation to understand a member or patient and route them to the right next step, while AI claims processing automates back-office decisions about payment and coverage. Intake is a lower-risk listening-and-routing function; coverage adjudication is high-stakes and, per the public record around UnitedHealth's litigation, is best kept with humans firmly in the loop. They often share infrastructure but carry very different regulatory exposure.

Why are health insurers replacing forms with conversational AI?#

Health insurers are replacing forms and phone trees with conversational AI because static intake flattens complex member situations into checkboxes and drives high call-center volume. A conversation can ask follow-up questions, capture context like a recent hospital discharge, and deflect routine requests while escalating sensitive ones. UnitedHealthcare reports Avery handles about 90% of member interactions without an advocate, illustrating the deflection and experience gains payers are pursuing.

Conclusion#

UnitedHealth Group and Optum show where conversational AI delivers first in a regulated, sensitive industry: at the front door. Avery's member-navigation rollout, backed by a roughly $1.5 billion AI investment, treats the first conversation about coverage and care as the highest-value AI surface — and keeps a human path open for the cases that need one. The contrast with the contested nH Predict litigation underscores the governance lesson: use AI to listen and route, keep humans accountable for what gets paid. For any team designing ai healthcare intake, the playbook is to start with the conversation, capture context a form never could, and escalate gracefully. That is exactly what Perspective AI's conversational intelligent intake is built for — start a study or see pricing to put it to work at your own front door.