Get Your Data House in Order Before You Buy AI

Every week, someone tells me some version of: "We need AI."

And every week, I ask the same question back: "Can you tell me how many customers you had last month, and would everyone on your team give me the same number?"

The pause that follows is the whole point of this post.

AI is the roof, not the foundation

I teach a framework called the Data House. Picture a building. At the very top is the roof: decision intelligence, automation and AI. That's the part everyone wants to buy.

But a roof sitting on cracked floors doesn't protect anything. It collapses. And when AI collapses, it doesn't fall quietly. It gives you confident, wrong answers, fast.

So before you buy the roof, walk the floors.

Not just a hunch, either. According to the U.S. Census Bureau's Business Trends and Outlook Survey, about one in five U.S. businesses reported using AI in early 2026, and use among firms with fewer than 20 employees didn't grow significantly between December 2025 and May 2026, while it rose at larger firms (U.S. Census Bureau). Bigger organizations tend to have more of these floors built. That gap is fixable.

The eight floors

Here's each floor, the crack that tells you it needs work, and one move you can make this month.

Floor 1: Governance

Policies, standards and ownership.

The crack: Nobody can say who owns a number, or which version is right. The move: Name an owner for each key metric and write down its definition. "Active customer" means what, exactly? Bought in the last 30 days? 90? Write it down.

Floor 2: People

Roles, skills and culture.

The crack: Knowledge lives in one person's head, and leaves when they do. The move: Find a data champion on each team and document what only they know. Every organization has a "spreadsheet whisperer." Protect that knowledge before they go on vacation.

Floor 3: Processes

Workflows and handoffs.

The crack: Work is marked "submitted" and quietly stalls between teams. The move: Follow one request from start to finish and record every handoff. You'll find the black hole. Everyone does.

Floor 4: Systems

The tools that hold the work.

The crack: Staff retype the same information into three different tools. The move: List every system you use and mark where the same data gets entered twice. Each double-entry is a place where mistakes are born.

Floor 5: Ingestion

How data comes in.

The crack: Paper forms, inboxes and loose spreadsheets are the front door. The move: Capture data digitally at the source, with simple checks at entry. A dropdown beats a free-text box. A required field beats a follow-up email.

Floor 6: Storage

Where data lives.

The crack: Files are scattered, duplicated and hard to find. "Final_v3_REAL_final.xlsx." The move: Consolidate into one home, with backups and a simple catalog that says what's where.

Floor 7: Processing

Cleaning and shaping.

The crack: Reports need hours of manual cleanup every month. The move: Automate the transformations you repeat, and test them. If you do the same cleanup every month, a computer should be doing it.

Floor 8: Visualization

Dashboards and reports.

The crack: Leaders wait days for a report they can't fully trust. The move: Pick the ten numbers that actually drive decisions and connect them live. Not fifty. Ten.

Then, and only then, the roof

Once your floors can carry weight, AI becomes powerful instead of risky. Your chatbot answers from data you trust. Your forecast is built on clean history. Your automation doesn't speed up a broken process; it speeds up a good one.

That's why I always say automation and AI, in that order, and only after the floors hold.

How to start this week

1. Score each floor from 1 to 5. Be honest. Do it with your team, not alone. You'll be surprised where people disagree.
2. Circle your lowest floor. That's where you start. Not with your favorite floor. The weakest one.
3. Make one move. Just the one listed above. Finish it before you start the next.
4. Revisit in 30 days. Score again. Progress you can see is progress you'll keep.

You can use the free Data House Checklist or take the two-minute Data House check right on my site.

Why I care so much about this

I've built data systems for global enterprises and for neighborhood nonprofits. The difference between the ones that work and the ones that don't is almost never the fanciest tool. It's whether someone did the unglamorous work on the floors.

Small businesses and community organizations deserve the same strong foundation the big companies have. The good news is that the floors don't require a big budget. They require attention, ownership and a little patience.

Pour me another: Which floor is your weakest? Tell me through the contact form, and I'll send you one extra tip for it.

Sip slow, stay curious.D