One file in. Each producer's own worklist out.

An address on your agency's authorization list emails the enrollment-platform export to a dedicated audit mailbox. The list is short and you control it. A producer who is not on it cannot start an agency-wide audit — that restriction is a property of the intake gate itself, not a setting layered on top of it. If a producer who is not on the list emails an export anyway, no audit runs, and the intake is built to tell your management list who sent it. The intake is built to refuse a sender who belongs to a different agency, rather than measure their file against the wrong agency's book.

The mailbox is checked on a repeating schedule, without anyone starting a job. A file you send is usually picked up inside a minute.

The attachment is read from bytes in memory. Caps on size, rows and columns stop an oversized file rather than letting it run.

If the file lost its name on the way through someone's downloads folder, VitalBot reads the header row and decides from the column names whether it is an enrollment export. Header names decide. A column count can corroborate; it never decides alone.

VitalBot reads the enrollment ledger your agency already keeps — the record of what your producers logged — and joins it to the export, row by row, on the application ID. Where an ID is blank it falls back to a name match and accepts one only at ninety percent confidence or better.

Everything it cannot join cleanly is separated out rather than guessed at: duplicate records on either side, IDs that are not usable, and clients who match no producer on your roster. Those become a review list for you, not a finding against a producer.

Columns are matched by their header text, not by their position, so an upstream reordering does not quietly change what is being compared.

Each discrepancy gets a name, a severity and an audience. Open and expired document requests, verification issues, cancellations that were never logged, policies marked unpaid that are live, and premium changes where a subsidy previously covered the whole bill are a producer's own work, and they go to that producer.

Findings your agency marks internal go to management only. Any finding can be marked internal, including the agent-of-record changes you did not initiate — that routing is your configuration's decision, not ours. Delivery to a producer requires an explicit entry in your configuration. A blank cell, an unrecognized value or a typo does not fall through to the roster: the finding is held back and flagged for you to look at. The default is to withhold.

Before anything is sent, four independent checks are built to hold the entire run: a flood of missing clients on one producer or across the roster, a column the audit depends on disappearing, the date anchor going missing, and a value-level flood where every column is intact but one code starts firing on row after row. A held run is diverted to management as a single digest naming the cause, and producers receive nothing.

A fifth check works one producer at a time, holding back that producer's findings while every other report goes out normally. Held is the safe state, and releasing a run is a human act. The section below sets out each of the five.

Each producer receives one HTML email: what needs attention, grouped worklists they can work straight down, then per-client detail ordered worst first. There is no spreadsheet attached. That is deliberate — a departing producer never receives an exportable client list from VitalBot. The spreadsheet version is written to the administrator's own disk and nowhere else.

On every completed audit, including a clean one, you and your management list receive one Agency Health Review: the counts, anything withheld from producers this run, and every internal cross-check that fired — plus the ones that did not, shown at zero, so nothing disappears quietly. It carries a button. Press send on the reply it opens, and the next scan mails back a spreadsheet of what each producer received on the most recent completed audit.

Every subject line states which date range of enrollments the audit covered. A send to a producer is recorded the moment it succeeds, so a run that resumes after an interruption does not email the same producer twice. Sending the export again starts a new audit, and producers who were already delivered will receive their report again.