Migrating your AMS? Here's what quietly disappears if you don't watch for it

I sat through a DOI audit in 2019 where the examiner asked for a producer's commission history going back six years. We'd migrated systems four years earlier. Two of those six years existed only as a PDF export nobody had opened since the day it was created. We found the numbers eventually, but it took three days and one very tense phone call with our old AMS vendor's support line, which by then had been sold twice.

That's the thing about AMS migrations. Everybody plans for the policies and the contacts. Nobody plans for the stuff that lives in the margins, and the margins are where the audit trail actually is.

Inventory the boring stuff, not just the obvious stuff

Before you sign a statement of work with any migration vendor, walk your current system and write down everything that isn't a policy record. I mean literally open every module. Most agencies inventory clients, policies, and carriers, then call it done. That covers maybe 60% of what's actually in the database.

The other 40% is the stuff built up over years of nobody documenting anything: activity logs tied to E&O defense, custom reports someone built in 2016 that finance still runs every quarter, attachment folders with signed applications, and every memo field a CSR ever typed a note into because there was nowhere else to put it. If you don't inventory it, your migration vendor won't either, because they don't know it exists. They're not being lazy. They're working from your data dictionary, and your data dictionary is incomplete.

Give yourself four to six weeks just for this step on an agency our size, 30 employees, maybe $12M in premium under management. Bigger shops need more. I've never seen an agency regret spending too long on inventory. I've seen plenty regret rushing it.

The three things every migration butchers

Commission history is the first casualty. Most AMS platforms store commission data as a rolling ledger tied to policy transactions, and migration tools are built to move policies, not ledgers. What often comes across is the current commission rate and maybe the last 12 months of payments. Everything before that gets "archived," which is vendor language for read-only and hard to query. If you need five years of history for a book valuation or a DOI request, confirm in writing, before cutover, that it's coming across as searchable data and not as a static export.

Custom memo fields are the second casualty, and they're the one owners underestimate the most. Every agency has fields nobody remembers creating: a "Special Instructions" field a producer used for renewal notes, a checkbox for "Do not auto-renew" that only three people know exists, a free-text field where a CSR tracked which clients required a signed EFT authorization. These fields rarely have a matching field in the new system, so the migration vendor either drops them or dumps them into a generic "notes" field where they become unsearchable. Either way, you lose the operational memory of your agency. Map every custom field to a destination field by name before migration starts, not after.

Attached documents are the third, and they're the one that will actually cost you money if it goes wrong. Signed applications, binder confirmations, loss runs, cancellation notices. These files are often stored with a database pointer rather than embedded, and a sloppy migration will move the pointer without moving the file, or move the file without preserving the folder structure that makes it findable. Six months after go-live, a CSR searches for a signed application and gets nothing. That's not a technical problem anymore. That's an E&O exposure.

Run a real parallel month, not a token one

Most agencies run a parallel period that's really just old system access left on for emergencies. That's not a parallel month. A real parallel month means every transaction gets entered in both systems for 30 days, and finance reconciles the two trial balances line by line at the end of it.

Do this for at least one full billing cycle, because that's the only way you catch commission timing errors, which are the single most common reason books don't tie after a migration. If your agency bills direct and also handles agency bill, run both cycles through the parallel month, not just one. I've watched agencies skip this because double entry is tedious and everyone's tired of the project by week three. That tedium is the whole point. It's where the discrepancies surface while you can still fix them against a live comparison, instead of six months later during an audit.

When the two systems tie for a full cycle, not just close, but tie, that's when you turn off the old one. Not before.

Build your data inventory and your parallel-month reconciliation plan before you sign the migration contract, not after.