Count how many places one customer's details live

Pick one good customer. Someone you have worked with for at least a year. Now count every place their details are stored in your business.

The accounting system has them, for invoices. The job or project system has them, for scheduling. The CRM, if you have one, has them. There is a spreadsheet the operations lead keeps because the job system's export is awkward. The sales lead's phone has the contact person's mobile. The email inbox has the last three addresses they wrote from. Someone uploaded last year's contract to an AI assistant to help draft a renewal, so now that tool has them too.

That example comes to seven. Your number will be different, and it is rarely one.

What goes wrong

Nothing, until something changes. Then the customer moves office and tells the person they deal with most. That person updates the copy they use. The other six copies stay as they were.

Next month the invoice goes to the old address and sits unopened. The technician drives to the old site. The renewal draft the AI produced greets a contact who left in the spring. Each of these looks like a small, separate mistake. They are the same mistake, made in seven places.

Ask which copy is correct and you usually get an answer that starts with "it depends who you ask." When the honest answer to "where is the truth" is "in somebody's head," the business has a reporting problem and a revenue problem, and is calling them admin.

Why adding a tool rarely fixes it

The natural reaction is to buy a system that will hold everything. Sometimes that is right. Often it becomes copy number eight, because the old places do not get switched off and people keep using the ones they are used to.

AI tools make the count worse in a quiet way. Every time someone uploads a document or pastes customer details into a chat to get help, a new copy is made, frozen at that moment, inside a tool that nobody else can see into. The tool will keep using it with full confidence long after it has gone stale.

One source per fact, written down

The fix is a decision before it is a system. For each kind of customer fact, the business names one place where the true version lives, and writes that down where everyone can read it:

• Billing address and payment terms: the accounting system.
• Site address and access notes: the job system.
• Contact people and their roles: the CRM.

Every other copy is a convenience copy. It can exist, but it is never the reference, and when it disagrees with the source, the source wins.

That short list belongs in the company's written memory, alongside the reason for each choice. Any AI tool that drafts customer-facing work should read the list first and pull customer facts from the named source, not from whatever was pasted into it last spring. And a person checks the customer details on anything that commits the firm, a contract, a quote, an invoice, before it goes out.

None of this needs new software. The list of sources can sit at the top of a shared document. What it needs is agreement, and a habit of fixing the source first when someone spots a wrong detail, rather than patching whichever copy they happen to have open.

What this does not solve

Naming a source does not keep it accurate. If nobody updates the accounting system when a customer moves, it is the wrong copy with an official title. Somebody has to own each source.

It also does not remove the copies already made. The uploads sitting in personal AI accounts, the old spreadsheets, the forwarded emails, they stay where they are until someone goes and clears them. That is tedious, and no tool does it for you.

Your ten minutes

Do the count for one customer, with whoever handles that customer day to day. Write down every place, and next to each one, whether it holds the current details. Then pick one fact, the billing address is a good start, and decide where its truth lives.

The RevOps audit looks at the same question across the whole path from lead to cash. Book 20 minutes at pgco.world/book.