Kueter says UHC voice agent handles more questions as models improve

UnitedHealth Group Chief Commercial Officer Dan Kueter, about five weeks into that role after more than four years leading UnitedHealthcare's commercial business, told a16z General Partner Julie Yu on Raising Health that much of the plan's back office work is already AI assisted, with the aim of reducing latency and human involvement. He described UnitedHealthcare's voice agent, Avery, as one of many in the market that keep gaining capability. As models improve, he said, so does the depth of conversation an agent can hold with a consumer and the range of questions it can resolve autonomously.

He went further on the clinical side. After conversations at the a16z Build event, he said agents are assisting physicians and in some aspects stepping in for them, stopping short of a definitive diagnosis, order or prescription but walking very close to that line. His conclusion: the technology will probably be ready before people's comfort is. In the same episode, per the Raising Health introduction, UnitedHealthcare is pushing to eliminate 30% of prior authorizations, and Kueter named touchless prior authorization as an internal objective.

For providers and wellness operators, this matters because the largest commercial payers are moving routine member contact to autonomous agents. Patients who get answers from a plan's agent in one call will bring that expectation to a practice's front desk, and staff who spend their day on repetitive calls stand to gain time for the conversations that need a person. Kueter's framing also puts the constraint where operators can act on it: not capability, but trust.

The evidence has limits. This was a podcast conversation, and no performance figures for Avery were given: no containment rate, resolution rate, escalation rate or satisfaction data. The observations about clinical agents were impressions from conference discussions, not published results. Treat the statements as a direction of travel from a large payer, not a benchmark.

The move: ask your contact center leader which call types could be resolved autonomously today, such as scheduling, benefits questions or directions, and which must stay with a person. Then decide how you would measure patient comfort alongside speed, for example by tracking how often callers ask for a human and how they rate the interaction. Building that measure before deployment lets you expand automation where patients accept it and keep staff where they add the most value.

You are an operations advisor to the COO and patient experience leader of [your organization], a [type of practice or health organization] that runs its own patient contact center.

Context: On the Raising Health podcast episode "The AI-Native Health Plan | UnitedHealth Group's Dan Kueter," UnitedHealth Group Chief Commercial Officer Dan Kueter said much of UnitedHealthcare's back office work is already AI assisted. He said its voice agent, Avery, can resolve a wider range of questions autonomously as models improve. He also said agents assisting physicians are walking very close to the line of diagnosis, and that the technology will probably be ready before people's comfort is. In the same episode, UnitedHealthcare described a push to eliminate 30% of prior authorizations. This was a podcast conversation. No performance figures for Avery were given: no containment, resolution, escalation or satisfaction data.

Task: Prepare me to ask our contact center leader, [name or role of your contact center leader], two things. First, which call types could be resolved autonomously today, such as scheduling, benefits questions or directions, and which must stay with a person. Second, how we would measure patient comfort alongside speed before any deployment, for example by tracking how often callers ask for a human and how they rate the interaction.

Output: A question set for that conversation, grouped under three headings: Call types and autonomy, Calls that stay with a person, and Measuring patient comfort. Add a short note on what baseline data we should collect first.

My inputs: our top call types are [your top call types], monthly inbound volume is [your monthly inbound call volume], our current share of callers who ask for a staff member is [your current rate, if known], and our phone platform is [your contact center platform].

Guardrails: Use only the facts above and the inputs I fill in. Cite the Raising Health episode by name. Flag anything uncertain or that needs verifying. Do not request or use patient data. Nothing here is medical advice. Label the result as a draft for the team to check.