Your data is messy. And it's not a technology problem.
If Tuesday's post resonated with you, here's the harder conversation.
Most small businesses have data scattered across spreadsheets, software platforms, and email threads. Customer names entered six different ways. Pricing that lives in one person's head. Processes that work because Mary has been doing it the same way for eleven years and everyone just knows to ask Mary.
That's not a data problem. That's a documentation problem.
And here's where it gets real: if you can't describe a process clearly enough to write it down, you can't use AI to improve it.
What data normalization actually means:
Imagine trying to pull every sale you've made to your best customer. Except half your team entered them as "ABC Company," two invoices say "ABC Co.," and one says "A.B.C. Company Inc."
The data is all there. But it's not usable.
That's a normalization problem. And it didn't come from bad software. It came from nobody writing down how to enter a customer name correctly.
That's just one version of this problem. The same thing is happening in your pricing, your inventory, your sales process, and your scheduling. Pick any core process in your business and you'll find a variation of the same story.
The root cause is almost always the same:
Someone learned the process, got good at it, and it never got documented. It just lived in their head. Then it got passed on informally to the next person, who added their own variation. Then the next person did the same.
Multiply that across every process in your business over five or ten years and you have a data environment that no AI tool can make sense of.
Garbage in, garbage out. AI won't tell you the output is unreliable. It will just give you a confident answer based on a flawed foundation.
So before you invest in AI tooling, ask yourself:
Can I describe my core business processes clearly enough to write them down?
If the answer is "mostly" or "it depends who you ask," you have work to do first. Not because AI requires perfection, but because the documentation process itself will reveal gaps and inconsistencies you didn't know existed.
That's actually valuable work regardless of what you do with AI afterward.
The good news:
You don't have to solve this all at once. Start with your highest-value processes. The ones that directly touch your customer, your revenue, or your competitive advantage.
Document those first. Clean that data first. Build from there.
There's a harder conversation underneath all of this, one about the people who hold that institutional knowledge and what they're thinking right now.
We'll get into that on my next post.