Does my company's data have to be clean before we start with an AI agent?
The short answer
No, not all of it. Only the data your first task depends on has to be in order. If you wait until every system is clean, you will never start. But an agent will not fix messy data for you. It will repeat the mess, and sound sure while doing it.
Why “clean everything first” never ends
Every company has old records, clients entered twice, and fields left half empty. A project to clean all of it has no natural end, and it competes with the daily work. Most such projects stall long before they finish.
The agent repeats what it finds
An agent answers from the data it reads. If a client appears twice, with two different addresses, the agent may use either one, and it will sound sure either way. Messy data does not make an agent hesitate. It makes it wrong in a convincing way.
Clean only what the first task touches
Pick the first task. Then ask three questions about the data that task needs:
- Is there one place where the correct version lives? If prices sit in your system and in three spreadsheets, decide which one is true, and let the agent read only that one.
- Is it current? Old contracts and old price lists should be marked as old, or kept out of what the agent reads.
- Can a person see where each answer came from? Then a wrong record shows up quickly, and gets fixed.
That is a few days of work for one task, not a year of work for the whole company.
Let the agent show you the mess
Once the agent works on real data, it shows you where the data is weak: the client it could not match, the price it found twice. Treat these as your cleaning list. Ask the agent to report what it could not settle, instead of guessing.
The question to ask your team
“For our first task, where does the correct version of each piece of data live?” If the answer is “it depends”, start there.