AI learns from the wrong data
We spent a decade cleaning operational systems for humans. Now AI needs the mess we filtered out.
These are the companion notes. The full essay lives on The Turing Pilgrim, my Substack publication.
Why this matters
A decade of digital transformation optimized operational data for human consumption — clean dashboards, reconciled reports, exceptions filtered out. That was the right call when the consumer was a person scanning a screen.
The uncomfortable turn: the mess we scrubbed away — the overrides, corrections, workarounds, and exceptions — is often exactly the signal AI needs to learn how the operation really runs, as opposed to how the system of record says it runs.
What this essay takes up
- Why data pipelines built for human dashboards make weak AI training grounds.
- What the filtered-out mess — overrides, exceptions, corrections — actually encodes about the operation.
- How to think about data readiness for AI without launching another multi-year cleanup program.
Who should read it
Anyone who has been told their AI initiative is blocked on “data quality” — and anyone writing the check for the cleanup. Read the full essay for the complete argument.
Related field notes
When AI Learns Physics…
Most industrial AI can notice patterns but cannot explain the physical story underneath them. Physics-informed neural networks change what the field can ask of a model.
June 11, 2026The Small Business AI Opportunity Is Hiding in Plain Sight
Small businesses are not short on ideas. They are short on capacity. AI works when it acts as a support layer around the owner — not as a replacement for their judgment.
May 22, 2026Why industrial AI needs its own org chart
IT is built for support. Operations is built for survival. Neither is structured to make AI work in the field — which is why industrial AI needs its own accountable operating model.
Working through this decision?
A short note on the AI bet you are weighing is enough to start. I usually reply within one business day.