All updates

Start with a small AI win

Paul Sponcia·November 20, 2026

Remember Dave Ramsey's debt snowball?

You pay off the smallest debt first, not the biggest one. It can look backwards if you are only looking at the size of each balance, but the point is momentum. That first payoff is something you can finish, and once you finish it you are more likely to keep going. The next one gets easier because you already proved to yourself that the method works.

AI works the same way. A lot of companies decide they need a giant, 18-month, change-everything project before they will start. A new platform, a plan for the whole firm, a design that has to look perfect on paper before anyone is allowed to try a real thing. I get the urge. It feels serious. It also means you wait a long time to see anything, and most people lose the thread before the big project ever lands.

This is where a lot of AI plans get stuck. Leaders want the first thing they build to be the thing they still run two years later, and that pressure makes the first thing too big. Then nobody wants to start, because starting now feels like signing up for a system you have to live with. If you give yourself room for the first build to be tuition, the first build can be small enough to finish. You may retire it, or you may keep it. Either way, your team has seen it, and that is what starts the next idea.

So skip the giant project, at least at the start. Start small. Pick one narrow thing a person on your team does often, build a small tool for it, and let people actually use it. Then do another, and then another. Each one is a small result, and those results stack. That is the snowball.

What do I mean by small? I don't mean a toy that nobody will touch. I mean a real use, just a narrow one, maybe a first draft that a person still reviews, or a recurring question that now has a first answer waiting. Keep it close to work people already do, show it, and let them react. If they shrug, you learned something cheap, and if they lean in, you have your next step.

And here is the part people don't like to hear. It's okay to build something that may not even be used a year from now. I mean that. If the first thing you stand up gets thrown away later, that does not mean you wasted the time. You showed people the power of what is possible, and their minds start clicking. Throwaway isn't wasted, it's tuition. You paid to teach the team what these tools can do, and that lesson stays even if the first app does not.

Once your team sees it, they start bringing you the ideas, and you are no longer the only person pushing. People closest to the work start saying, what if we tried this, or I already tried a rough version at my desk. The snowball isn't you forcing a program down the hall, it's the team leaning in because they have seen a real result, even a small one, and they want the next one.

If I were sitting with an owner, I wouldn't start by laying out the 18-month plan. I'd ask for one small use that would make a normal week easier, something you could try soon, something you would be willing to throw away if it taught you something. Then I would want the team to see it, even if it is rough. That first look is what gets people thinking.

This is also how AI enablement should start, with a short list of real work and a first result people can feel. If that first result takes a slow, repeat task off someone's plate, workflow automation is often a good kind of small start. The wider AI and automation work can come after the team has seen what good looks like.

What is the smallest AI use you could try in the next few weeks, and would you be willing to throw it away if it taught your team something?

Let's talk

Let's give your team their time back.