“We can’t use AI properly until we fix our data.”
I keep hearing this from CSCOs and Ops VPs. And it is becoming the default excuse for not starting, for failed pilots, for staying comfortable with the status quo.
Data quality is a real problem. Nobody is denying that.
But here is the thing: your data is already running your business.
You close the books every month on it. Your board reviews financials built from it. Customer promises get made on it.
If the data is so broken that AI cannot touch it, how is the business still operating?
It is not broken. It is imperfect. There is a difference.
Data will never be perfect. Waiting for it means waiting forever.
What works is prioritising.
Pick the decision you want to improve, find the data slice that drives it, and start there.
- If expedites are eating margin, start with the top lanes and SKUs driving premium freight and model what would have prevented it.
- If inventory is bloated, start with the top 50 SKUs where service targets and reorder policies are driving working capital.
You do not need perfect data across the full supply chain. You need the right data for the right question.
The data problem is rarely why AI does not work.
It is why organisations stop before they learn what works.