When AI enrichment writes the wrong country.

Somewhere in your CRM settings there is a toggle offering to fill in missing fields automatically. It reads like free labor. Most admins turned it on the week it shipped and have not thought about it since.

We have watched what that toggle can do, inside more than one client instance. One of the most widely used sales platforms in the world runs native AI enrichment, and on state and country fields we kept catching it getting companies wrong. Roughly 10 to 15% of the accounts it touched, in the instances where we measured it.

Ten to fifteen percent may not sound like a crisis. Run it against a sophisticated territory plan and it becomes hundreds or thousands of accounts assigned wrong, routed wrong from marketing, credited wrong in comp, and quietly baked into the next planning cycle. At one client the problem announced itself through the territory map. Territories that had been planned outside the CRM started drifting out of alignment: some grew considerably, others shrank, and then reps began noticing that some of their best accounts were no longer theirs. That was the toggle running for a couple of days. Loose on the full CRM, it would have undone the planning work that had gone into that map.

What happens when AI enrichment writes the wrong state?

The record is the least of it. Geography sits underneath territory assignment and lead routing, so a wrong state does not stay a wrong state. It becomes a lead routed to a rep who will never work it, an SLA burning against the wrong queue, a territory dispute between two sellers with equally defensible claims, and a regional pipeline report segmented by fiction.

Follow the field one step further and it turns into money. Territories feed comp plans. When an account lands in the wrong region, the wrong rep gets credit or the right rep gets nothing. Comp disputes nobody can adjudicate follow, because the field everyone would use to settle the argument is the field the machine rewrote. Attainment conversations get harder from there, and attainment was already in trouble.

The quietest casualty is the account that lands nowhere. Geography that matches no territory leaves accounts unassigned, and an unassigned account is a gap in the market you believe you cover, with no rep on it and nobody assigned to notice.

The blast radius is the system, not the field

A bad write in a foundational field propagates through every rule keyed on it, which in most B2B instances means routing, assignment, scoring, reporting, and comp. According to Salesforce’s State of Sales, 7th edition, a survey of 4,050 sales professionals across 22 countries published in February 2026, 87% of sales organizations use some form of AI, and 74% are actively doing data cleansing to improve AI returns. Read those together: the same organizations turning AI loose on their data are running cleanup crews behind it.

The routing engine does not know the state is wrong. The scoring model does not know. The dashboard your board reads does not know. Each layer treats the field as settled fact, because that is what CRM fields are supposed to be, and the error passes silently through all of them until a rep raises a hand about an account that cannot possibly be theirs.

Why does AI enrichment get geography wrong?

Because it infers instead of verifies. Companies share names, entities span countries, headquarters and subsidiaries blur, and a model completing a blank field writes its best guess with the same confidence it writes a certainty. Nothing in the workflow asks it to show its work.

The gap between how ready companies feel and how ready their data is has been measured. Per the 2026 State of Data Integrity and AI Readiness from Precisely and Drexel University’s LeBow College of Business, a survey of 505 data and analytics leaders, 87 to 88% claim their organization is AI-ready while 43% admit data readiness is their biggest obstacle. Those are the same people answering the same survey. And a Harvard Business Review Analytic Services survey of 1,574 enterprise IT leaders, sponsored by Cloudera and published in March 2026, found just 7% say their data is completely ready for AI. Enrichment features ship into that gap and write into it.

A human who did this would be managed

If a new hire quietly re-keyed the geography on a meaningful share of your accounts, you would catch it inside a week and there would be a conversation. When an enrichment feature does the same thing, the writes arrive one field at a time, look authoritative, and belong to nobody. There is no name attached to review, no pattern a manager is watching for, and no one whose job description includes noticing.

That is the structural difference between a bad hire and a bad automation. The hire has a manager. Give every automated writer in your CRM one too: a named human who owns what it writes, reviews samples of its output on a cadence, and can turn it off without a change request.

Machine writes need a quarantine

Enrichment can be useful. The version that survives contact with a revenue system follows four rules. Machine-written values land as suggestions, never silent overwrites, on any field that routing or comp reads. Every write carries provenance: what wrote it, when, from what source. The named owner sample-audits the output on a schedule. And geography, ownership, lifecycle stage, and anything a comp plan touches require a human confirm before the value goes live.

Your sellers already spend most of their week not selling. Assigning them to re-verify a machine’s confident guesses, one support ticket at a time, is a strange way to spend the hours that remain. When we run our free CRM diagnostic, the reachability and field hygiene checks surface this class of problem quickly, and the connection is read-only, so nothing writes anything.

Someone will decide what is true in your CRM

Go find the toggle. Ask who owns what it writes. If the answer is a vendor’s roadmap, you have delegated your territory map to a feature, and features do not sit in comp reviews.

Someone is going to decide what is true in your CRM this quarter. Choose who.