All posts

Insurance

Insurance open enrollment: AI voice agents as seasonal capacity, not seasonal cost

NeuraVoice··11 min read

The week of November 15 at a mid-sized independent agency in Tampa: phones ringing 4x baseline, three CSRs on back-to-back Medicare AEP calls, two callers in queue for over nine minutes, a Marketplace prospect on hold who hangs up at the eight-minute mark, and the principal agent fielding overflow she has not personally taken in six years. Two weeks earlier, the same office handled the same call volume with the same staff and nobody noticed. That is what open enrollment does to insurance operations.

The reflex answer for the last twenty years has been to hire seasonal CSRs. The cost of that reflex has gotten worse, not better, and the math now favors a different category of capacity entirely.

Open enrollment is not one season, it is four overlapping ones

Most outside the industry think of "open enrollment" as a single calendar window. Inside an insurance agency, it is four overlapping cycles, each with its own compliance posture and call-volume curve:

  • Medicare Annual Election Period (AEP). October 15 through December 7. Federally fixed dates, CMS-regulated marketing, and the highest per-call compliance overhead of the four. This is the volume-driver for any agency with a Medicare book.
  • Marketplace Open Enrollment Period (OEP). November 1 through January 15 in most states, with state-based exchanges sometimes running longer. ACA subsidy questions, special-enrollment qualifying events, and a different prospect demographic than AEP.
  • Employer-sponsored open enrollment. Concentrated in Q4, typically October through early December, with smaller spikes in January and at fiscal-year boundaries for non-calendar plans. Group-benefits agencies feel this one.
  • Medicaid recertification. State-by-state cadence, mostly post-PHE redetermination on rolling 12-month cycles. Less concentrated than AEP but a steady drag on call capacity year-round, with state-specific spikes.

Three of the four overlap in November. An independent agency with a mixed book (Medicare, individual ACA, group benefits) is fielding all three streams simultaneously, on the same phone lines, with the same CSRs.

The volume math is sharper than agencies usually quote it

A typical mid-sized independent that handles 200 calls per week year-round will see 800 to 1,000 in peak AEP weeks. The breakdown is not uniform:

  • T-65 inquiries (people aging into Medicare) compress into a narrow window.
  • Existing Medicare Advantage members calling about plan changes, formulary updates, and PCP network shifts.
  • Marketplace renewals and subsidy re-certifications.
  • Carriers and SOA forms that need to be reviewed and signed before any plan-specific conversation can legally happen.

The average call length goes up too. A routine billing question in March is six minutes. A Medicare plan-change conversation in November, with required disclosures, scope-of-appointment review, and plan-comparison walk-through, runs 22 to 35 minutes. So the agency is not just fielding 4x the calls, it is doing it with a longer per-call burn. Effective capacity demand goes up 6x to 8x against a CSR pool that grew 0%.

Hiring three seasonal CSRs costs more than the line item suggests

The traditional answer is to hire two to four seasonal CSRs in September, train them through early October, run them October 15 through January 15, and let them go. The line-item budget is wage cost. The full-loaded cost is not:

  • Recruiting and screening: 60 to 90 hours of agency-principal time at the worst possible moment in the calendar.
  • Pre-AEP compliance training: AHIP certification (~$175 per person), carrier-specific certifications for every contracted plan (Humana, Aetna, Cigna, UnitedHealthcare, Wellcare, Anthem each maintain their own), CMS marketing rules, state DOI rules, and the agency's own SOA workflow.
  • Tools and licenses: agency management system seat, dialer license, e-SOA tool, screen-recording compliance tool, Medicare quoting platform.
  • Productivity ramp: a brand-new CSR is at 40-50% of an experienced CSR's effective throughput for the first three weeks, which is most of October.
  • Carrier appointment lag: in some states, a newly-hired CSR cannot legally discuss plan specifics until the carrier appointment is processed, which is 10 to 30 days.
  • Lost-knowledge cost: in mid-January you let them go, and any carrier-specific edge cases they learned walk out the door with them.

A reasonable all-in number per seasonal CSR, including wages, certifications, tools, training time, and ramp inefficiency, is $40,000 to $70,000 for a 14-week engagement. For an agency hiring three, that is $120,000 to $210,000 of seasonal capacity that vanishes on January 16.

AI voice capacity scales on a different cost curve

The economics of an AI voice agent are not a per-CSR fixed cost. They are usage-based, and the relevant properties for seasonal capacity are different:

  • Elastic to volume. A 4x or 10x call-volume spike does not require hiring. The same agent configuration handles 50 concurrent calls or 500.
  • 24/7. Open enrollment callers do not respect 9-5. A T-65 caller in Phoenix at 7:40 p.m. local time, who reaches voicemail today, reaches a structured-intake conversation instead.
  • No training drift. A CSR who has heard the same Humana PPO question 200 times by November 20 starts to drift on the disclosure language. The agent does not.
  • Specialty-routable. Medicare AEP, Marketplace OEP, and group-benefits inquiries each have different schemas, different compliance posture, and different downstream handoffs. A well-built agent routes them at the qualifying-question stage, captures the right fields, and hands off to the right human.
  • No mid-January cliff. The agent is still there February 1 doing claims-status triage and after-hours overflow.

The cost shape is closer to a utility bill than to a payroll line. For an agency that paid $150,000 in seasonal wages last year to absorb the AEP spike, comparable AI voice capacity over the same window typically lands well below that, and the residual capacity stays useful year-round.

Medicare adds a compliance layer most vendors underestimate

Medicare AEP is not just a volume problem. It is a compliance-regulated marketing channel governed by CMS, and the rules tightened materially in the 2024 Medicare Advantage and Part D Final Rule (CMS-4201-F). Specifically:

  • Third-Party Marketing Organization (TPMO) disclaimers. 42 CFR § 422.2267(e)(41) requires the standardized TPMO disclaimer on inbound and outbound marketing calls discussing MA plans. The disclaimer must be read in the first minute. An AI agent doing AEP intake has to read it.
  • Recording requirement. All marketing, sales, and enrollment calls must be recorded and retained for 10 years (42 CFR § 422.2274(g) and parallel Part D rules). The agent has to record, the agency has to retain.
  • Scope of Appointment (SOA). A documented SOA must precede any specific plan discussion. The agent can capture and timestamp the SOA before transferring to a licensed agent.
  • Marketing vs. non-marketing distinction. CMS distinguishes "marketing" calls (subject to the full ruleset) from "communications" (lighter ruleset). Whether a given call is marketing depends on content, not labeling. The agent's prompt design has to know the difference.
  • Inbound vs. outbound. Inbound calls to a published number are treated differently than agency-initiated outbound. TCPA still applies on the outbound side, and AEP outbound calling rules add a layer on top.

A vendor who has not built specifically for AEP will hand you an agent that violates one of these on the first call. The right vendor question is which CMS provisions the agent is configured against, line by line.

For TCPA-side context on AI voice and consent, see TCPA AI voice and legal intake (the TCPA framework applies to insurance outbound the same way). For the recording requirements, state call recording laws and AI voice agents covers the two-party-consent state list every agency book crosses.

State DOI rules force an AI/human handoff design

A common failure mode is to treat the AI agent as a replacement for the licensed agent. It is not, in any state. State DOI regulations vary, but the consistent principle is that a licensed, appointed agent must be involved at specific call points: plan recommendation, application submission, and certain disclosures. The clean design pattern is:

  1. AI handles the inbound, identifies caller intent, captures qualifying information (eligibility status, ZIP code, current coverage, drug list, doctors, scope of appointment).
  2. AI reads required disclosures and confirms SOA.
  3. AI either schedules a callback with a licensed agent on a specific calendar slot, or warm-transfers in real time to an available licensed agent on the floor.
  4. Licensed agent does the plan recommendation and enrollment.

The handoff itself is where a lot of vendor demos fall apart. A blind transfer that drops context loses the call. A warm transfer with structured intake delivered to the licensed agent's screen before the call lands is the working pattern. The mechanics of this are covered in warm vs blind live transfer.

Independent agencies benefit more than captives

Captive agencies (a single carrier, single product schema, single compliance posture) get a smaller marginal lift from AI voice. The complexity of their intake is bounded. An independent agency with 12 carrier appointments and 40+ active plan options has to route every call by carrier, plan type, eligibility, and state. That routing is exactly what an AI agent does well. The complexity penalty an independent CSR pays for the breadth of the book is the complexity premium an AI agent capitalizes on.

The other dimension is contract leverage. Captive agencies have less flexibility on tooling decisions because the carrier dictates much of the stack. Independents pick their own AMS, dialer, and intake layer. AI voice fits into that stack without carrier sign-off.

Five vendor-evaluation questions for seasonal capacity in insurance

Before signing anything for AEP 2026, ask the vendor to answer these on a live call, not in a deck:

  1. Show the TPMO disclaimer in the agent's prompt and the call point at which it is read. Confirm it complies with 42 CFR § 422.2267(e)(41) language requirements.
  2. Walk through the call recording and 10-year retention architecture. Where is the recording stored, who has access, and what is the audit trail?
  3. Demonstrate a live SOA capture with timestamping and storage in your AMS. Can a licensed agent retrieve it before the warm transfer?
  4. Show the routing logic for a multi-carrier independent book. How does the agent distinguish a Humana MAPD inquiry from an Aetna SNP inquiry from a Cigna PDP inquiry, and route each to the right specialist?
  5. What is the fallback when the agent does not know the answer? "Escalate to a human" is not specific enough. The pattern should be a defined escalation path with structured context delivered to a human who can answer in real time, not a callback request that recreates the original problem.

If a vendor cannot answer all five with screens and call recordings, they have not built for AEP. The pricing question (per-minute versus flat) only matters after they pass the five.

The contrarian read

The interesting cost in seasonal CSR hiring is not the wage line. It is the opportunity cost of agency-principal attention in September and early October, which is the same window in which the principal should be doing pre-AEP outreach to the existing book. Every hour spent recruiting and onboarding a temp CSR is an hour not spent on a renewal conversation with a member who is at risk of switching carriers. The seasonal-CSR model does not just cost money. It costs retention.

The frame of AI voice as "seasonal capacity, not seasonal cost" is real, but the deeper frame is that it returns October back to the agency principal. That hour is worth more than $40,000.

For the cost-comparison side of this, see AI voice pricing per-minute vs flat. For the warm-transfer mechanics that make the AI/human handoff work for licensed plan recommendations, see warm vs blind live transfer. For the recording-law foundation underneath the CMS retention rule, see state call recording laws.

If you run an agency book that crosses Medicare, Marketplace, and group benefits, AEP 2026 starts in roughly four months. The decision about how to staff it is being made now.

Book a call to walk through what AEP-ready intake looks like on your specific carrier mix, start free trial, or see the pricing page.

Keep reading

Try it on your own calls

Spin up an agent in 5 minutes. Cancel anytime.

14-day free trial, 60 free minutes, cancel anytime. Bring your own intake script or start from a template. Wire it up to your CRM when you're ready.

Start free trial

Want a guided walkthrough first? Talk to the team.