Last updated: September 12, 2026
What to check each week of AEP
An AEP call quality plan sets what gets checked each week of the enrollment window that runs October 15 through December 7. Week one confirms the disclaimer lands before benefits on live calls. Later weeks track benefit accuracy, Scope of Appointment capture, and the enrollment confirmation as call volume climbs.
Two dates set the shape of the next eight weeks. The CY2027 marketing provisions took effect on October 1, so calls this season get scored against the current rules rather than last year's. Enrollment opens October 15 and closes December 7, which is 54 days and a little under eight weeks of live volume.
Working backward from the 15th, the week you are in right now is the last one where a script change is cheap. Phone paths, retention settings, and the script review itself belong to that week, and they live on CMS call recording requirements rather than here. Everything below starts on the 15th.
Week one, the first five days
Week one carries more weight than the seven weeks after it, which is our read rather than a measured finding. New seasonal agents are on the phone for the first time, the script is the newest thing in the building, and the habit that forms in the first five days tends to hold through December.
Start with disclaimer position. The TPMO disclaimer is conveyed verbally during sales calls prior to the discussion of any benefits under 42 CFR 422.2267(e)(41)(ii), which makes it a fixed phrase with a trigger that moves around inside the conversation. Pull five calls for each newly licensed agent, mark the second the disclaimer starts, mark the second a benefit is first named, and compare the two. The rule behind the check sits on the TPMO disclaimer timing.
Sampling method is the second check. A random pull across the floor leans toward veterans, because veterans take more calls, and the people you actually need a read on have never worked an enrollment season. Pull by agent instead of by volume, and make sure every seasonal hire has at least one scored call by day five.
Third is a process check on the supervisors. An auto fail that sits in a queue until the monthly review does nothing for you in week one. Set the rule that anything flagged as a critical fail gets opened the same day it flags, then on Friday go look at whether that actually happened.
None of this is a defensible sample and we would not pretend otherwise. Five calls per new agent, read closely on day three, beats 300 calls read in January, which is most of the argument for running the check at all.
Weeks two through four, as volume climbs
Benefit accuracy is the week two check, or it is if your season opens on inbound. An agency pushing outbound off a mail drop should move the consent item up to week two and let benefit accuracy wait until three. Either way, what you are scoring is whether the benefit language traces back to the plan's own approved materials, whether the plan year got named, and whether any comparison the agent made is one those materials support. The content prohibitions sit at 42 CFR 422.2262.
Somewhere in the same stretch, pull the Scope of Appointment paperwork against the calls. An SOA lists which product types the beneficiary agreed to discuss, and the failure worth scoring is a conversation that drifts into a product type the form never covered. Cite the requirement from 42 CFR 422.2264 rather than from a training deck.
Consent scope on outbound is the one most likely to bite. A consent captured for one product line does not carry over to another on its own, and where the line falls is a question for your counsel rather than for a scorecard criterion.
Weeks five through eight, the close
By the middle of November the volume flattens out and the failures move to the end of the call. Enrollment confirmation is what to score: whether the agent confirmed the beneficiary understood what they were enrolling in before the application went in, and whether the required next steps were actually given rather than assumed.
One item in this stretch has nothing to do with CMS. Before any of these scores get used in a coaching conversation or a plan partner report, two supervisors should score the same five calls separately and compare. Calibration variance reads as a trend when nobody checks for it, and a floor that looks like it got worse in week six is sometimes a second reviewer who reads the disclaimer criterion differently. QA calibration for a call center has the session format.
The checklist, in one table
Nine checks across the window, with an owner and an evidence column so the thing can be handed to somebody instead of living in a supervisor's head.
| Week | Check | Criterion source | Owner | Evidence required |
|---|---|---|---|---|
| Week 1 | Disclaimer starts before the first benefit is named | 42 CFR 422.2267(e)(41)(ii) | Compliance lead | Transcript timestamps for the disclaimer and the first benefit mention |
| Week 1 | Every seasonal agent has at least one scored call by day five | Internal QA policy | QA supervisor | Scored call list by agent, not by call volume |
| Week 1 | Critical fails opened the same day they flag | Internal QA policy | QA supervisor | Flag date and review date on each auto fail |
| Weeks 2 to 4 | Benefit language traces to approved plan materials | 42 CFR 422.2262 | Compliance lead | Scored criterion plus the material the language came from |
| Weeks 2 to 4 | Plan year and effective date named on the call | 42 CFR 422.2262 | QA supervisor | Transcript excerpt with the plan year stated |
| Weeks 2 to 4 | Scope of Appointment covers every product type discussed | 42 CFR 422.2264 | Sales manager | SOA record matched against the product types in the transcript |
| Weeks 2 to 4 | Consent scope on outbound matches the product line called about | Counsel guidance, TCPA | Compliance lead | Consent record with its source and date |
| Weeks 5 to 8 | Understanding confirmed before the application is submitted | Internal QA policy | QA supervisor | Scored criterion with the confirmation timestamp |
| Weeks 5 to 8 | Two supervisors score the same five calls and compare | Internal QA policy | QA manager | Variance sheet with the resolved reading noted |
The same nine rows download as a CSV with no form in front of it.
What this looks like when you are scoring every call
Everything above is a sampling exercise, because that is what a supervisor's calendar allows. Say you run 18 agents at 35 calls a day. The floor takes roughly 34,000 calls across the window, and one to two reviews per agent per week, the volume commonly assumed for agency QA and a figure that comes from client research rather than anything we measured, gets you about 290 of them. Under one percent.
Most vendors in this category quote some version of that ratio, so treat coverage as the setup. What changes when the scorecard runs on the whole volume is where the supervisor's hour goes. The weekly check stops being "pull five calls and listen" and turns into "open the 14 calls the system flagged".
CallSniper ingests calls after they end, transcribes them, separates the speakers, then scores against a scorecard you define. The disclaimer criterion becomes a phrase set in the vocabulary library rather than something a reviewer listens for, and a call where a benefit was named first can be set as a critical fail, which puts it on somebody's desk on October 16 instead of in January. Each flag carries a timestamp and a confidence score, so the low confidence ones get the human minutes. A Medicare sales call scorecard holds the criteria and the auto fail flags behind every check on this page.
Two limits belong here. Rule based criteria score reliably and tone scoring gets shakier, which Aircall and StackAI both note about automated QA generally, so whether an agent sounded pushy stays a human read. The bigger one is the script. A checklist run perfectly against wording that was wrong on October 1 tells you it was wrong on every call, which is useful and also late, and the fix lives in writing a compliant Medicare call script.
Audio stays on Trinsic's own infrastructure and runs against Trinsic's own model, so a recording carrying a beneficiary's name and plan details never goes out to a public model provider. More on call QA for Medicare agencies.
Frequently asked questions
When does AEP start and end?
The Annual Enrollment Period runs October 15 through December 7, which is 54 days, and the changes made during it take effect January 1. Confirm the current year's dates on cms.gov before building a calendar around them, since the enrollment calendar is the one thing here nobody should take from a blog post.
How many calls should you audit during AEP?
Most agencies work from one to two calls per agent per week, a figure that comes from client research rather than a published standard. Size it against your own volume instead. At 18 agents and 35 calls a day that ratio covers well under one percent of the season. Decide what you want to be able to prove afterward.
What should you check first during AEP?
Disclaimer position, starting on day one. The disclaimer has to land before any discussion of plan benefits, it is fixed wording with a trigger that moves, and comparing the second it started against the second a benefit was first named is arithmetic once the call is transcribed with timestamps.
If you only have an hour in week one
Nobody runs a full QA program in the first week of enrollment. Two checks catch most of the risk per minute spent, and a third is worth it if the hour holds.
Disclaimer position on five calls for each newly licensed agent, which at four minutes a call is 20 minutes per new agent. A failure there is unambiguous and mechanical, and it is the only item on this page where the answer is a number rather than a judgment.
The auto fail queue from day one takes about ten minutes to read and tells you whether anybody is working it. Then, if there is time left, check whether any agent has zero scored calls by Friday, because the agent nobody has listened to is the one you hear about in December.
A call review on six months of archived audio gives you the same disclaimer number for last season while this one is still running, at a fixed pilot price.
Put the week one check on the calendar for October 16 with a name next to it. A check that lives in somebody's head does not survive a 600 call day.
Sources
- Annual Enrollment Period running October 15 through December 7. Standard CMS enrollment calendar, widely established. Confirm on cms.gov or medicare.gov before publishing.
- TPMO disclaimer timing, 42 CFR 422.2267(e)(41)(ii), read via Cornell LII on 2026-09-09. Page shows most recent amendment 91 FR 17583, April 6, 2026. Verified against a primary text mirror. Read it again on eCFR before publishing.
- Marketing content prohibitions, 42 CFR 422.2262. Not read during drafting. Unverified, confirm before publishing.
- Scope of Appointment requirement, 42 CFR 422.2264. Not read during drafting. Unverified, confirm before publishing.
- October 1, 2026 effective date for the CY2027 marketing and communications provisions (CMS-4208-F3). Secondary, from agent facing summaries gathered 2026-09-08. Verify against the Federal Register publication before publishing.
- TCPA consent scope on outbound calls. General, unverified, routed to counsel in the body.
- One to two calls per agent per week as typical agency QA volume. Client research from 2026-09-08, not independently measured.
- Weekly check cadence and the weight placed on the first five days. Editorial judgment, offered as a recommendation rather than a finding.
- Aircall and StackAI, on AI call scoring being more reliable on rule based criteria than on subjective tone. Third party, gathered 2026-09-08. Unverified.