In one sentence
An identifier is a stable, unique label that lets you refer to one specific thing without confusion.
The simple model
A customer number, an order ID, an invoice number: each of these points to exactly one thing and does not change. That is an identifier. It lets people and software say "this exact record" without relying on a name that might be shared, misspelled or changed.
Names are not identifiers. Two customers can be called "John Smith". A company can be written five different ways. An identifier removes that ambiguity.
Why you care
Identifiers are what let records be matched across systems and over time. Without them, workflows quietly break in familiar ways:
- The same customer appears twice because nobody could tell the records were the same.
- An order cannot be linked to its invoice because there is no shared reference.
- An update lands on the wrong record because two had similar names.
For AI workflows this matters because matching is often the hidden hard part. An AI can extract a customer name from an email, but linking that name to the right record needs a reliable identifier, or a careful matching step, underneath.
Good identifiers are stable and unique
A useful identifier does not change (a customer keeps the same number even after moving), and no two things share it. Reusing or changing identifiers causes exactly the confusion they were meant to prevent.
A small example
A company matched invoices to orders by supplier name. Slight spelling differences meant hundreds of invoices could not be matched automatically and were handled by hand. Introducing a shared order reference on both sides let the matching happen automatically, and the manual work largely disappeared.
What this does not solve
An identifier makes records matchable; it does not make the underlying data correct. Two records can share an identifier by mistake, or point to outdated information. Identifiers remove ambiguity, not the need for accuracy.
