Before AI, Clean Your Data: The One Thing to Do This Week

By Luba Inz ·

Clean up your data. Check your sources. When an owner asks me for one thing to do this week, that's my answer, and it stays my answer when the question is AI. Clean data before AI, every time. It's the most painful, hardest step, and it's the one that opens your eyes.

Beautiful reports, and your gut says no

Management can give you beautiful reports, show you beautiful numbers, even tell you everything is great. And your gut is telling you: no, it's not.

When that happens, the root cause is always the same: how correct and trustworthy your data is.

At a home-services company where I ran operations, everyone believed organic was the cheapest lead. So I calculated the real cost of organic leads: the repeat rate, the average ticket. In some locations, the cost of mistakes and bad reviews meant organic cost almost more than paid. Its ROAS was lower than planned and hard to track.

Where your data really comes from

This week, go check how your data gets recorded. Is it trustworthy? Is it updated? Is it usable? Is every department tracking?

Sorry, but I make everyone report. Without documented proof of what you do at work, I don't know how to manage you. I always need to see where your data comes from, so I can trust it.

That same company is where clean data paid off. The retention department was the first one we automated with AI, in 2023, when few companies were doing it. We ran AI-powered analytics on our customer database. AI lead scoring found the best-fit customer clusters, and those clusters became the most profitable. The key was clean, verified data first.

If you're not sure where your company stands, start with is my company ready for AI?

The checklist for this week

Go through every department and ask four questions about the data it produces:

  • Is it trustworthy? Where does each number come from, and can you see the source?
  • Is it updated? When was it last recorded?
  • Is it usable? Can someone work with it, or does it need fixing first?
  • Is every department tracking? Or are some running on memory?

It doesn't sound easy-peasy. But that's where you always start.

Then: write the processes that repeat

If it's about AI, the second step is to ask your team: which processes repeat? If they're not written down, write them as SOPs. The SOP is the foundation for automation.

A good SOP covers:

  • the order of steps
  • where data comes from and where it's saved
  • the KPIs to track
  • an owner for each step

You can see what that looks like in an SOP written for AI agents.

Keep a responsible person who owns the outcome. Give them the tools to add their own agents, but by the SOP.

Then pick the most repetitive, most time-consuming process and automate it together with the person who owns the outcome. What took them hours now takes minutes, and they'll be grateful. Everyone agrees on how it should be done. If you're choosing where to begin, here is which processes AI can take over now.

The mistake I see owners make is caring only about results, not about how the team works. Care about how your team makes your company stronger. If a tool is clunky or eats their time, change it.

I've worked with owners who only cared about results. Then people came back with: "I couldn't log in to the tool you make me report in. It eats 5 hours of my week and hurts my work." You have to listen. You gotta listen.

The CEO's job is to guide and mentor, hire good people and help them. I believe that wholeheartedly. Writing SOPs together with your team and helping them build an effective process is what separates a good COO, CEO or manager from a bad one.

Start with the data

Efficiency Sprint makes companies operationally ready for AI. The order doesn't change: clean data, then the processes that repeat, then the agents.

Want to know where your company stands today? Take the AI readiness quiz.