In one sentence
A system of record is the official system that owns and maintains a particular type of data.
The simple model
Companies run many systems. One holds customers, one holds orders, one holds staff. The system of record for a data type is the one that is treated as its official home. Customer details belong to the CRM, orders belong to the order system, and so on.
A source of truth is about a specific fact. A system of record is about a whole category of data. Often the system of record is the source of truth for the facts it owns, which is exactly the tidy situation you want.
Why you care
When a workflow needs to read or change data, it should go to the system of record for that data. Reading a customer's address from a random spreadsheet, when the CRM owns customer data, is how workflows end up acting on stale information.
For AI workflows this is a design decision worth making explicit: which system does each step read from, and which system does it write to? See read access vs write access.
Watch for shadow systems
A "shadow system" is an unofficial store that has quietly taken over part of a system of record's job, usually a spreadsheet. It happens when the official system is hard to use. It is not wrong to notice this, but automating on top of a shadow system builds on sand. Understand why it exists before deciding whether to keep or replace it.
A small example
A workflow updates delivery addresses. There is a CRM (the system of record for customers) and a spreadsheet the dispatch team edits. Addresses are changed in the spreadsheet and never make it back to the CRM. Any automation reading the CRM gets old addresses. The fix is to make the CRM the place addresses are changed, or to sync the two deliberately.
What this does not solve
Knowing the system of record does not mean it is easy to connect to. Some systems have no clean way for software to read or write them. That is a real constraint you will meet in the digitize stage, through integrations and APIs.
