Point Your AI at One Lever, Not Everything

Part three of three. You don’t pull 45 levers. You find your weakest bucket, pick one, and test it in 30 days.

Over the last two issues I laid out the four buckets. Cost savings. New customers. Bigger customers. Loyal customers. By now you might be staring at all four and feeling the pull to fix every one of them at once.

Don’t. Trying to fix everything is the fastest way to fix nothing.

I keep a playbook of 45 growth levers across those four buckets. I’ve pulled most of them inside real businesses, so this isn’t theory. You won’t pull 45. You’ll pull one. The whole game is picking the right one, and that starts with finding where your business leaks the most.

So here’s the test. Walk each bucket and answer one honest question.

New customers: do leads come in and then sit? If you’re slow to respond and deals go cold, that’s costing you.

Bigger customers: when did you last raise your prices? If the answer is “a while back,” you’re leaving the easiest money on the table.

Loyal customers: do people buy once and quietly drift away? If you don’t even know your repeat rate, that’s a leak you can’t yet see.

Less friction: is your team burning hours on work a machine could do? If the same manual task eats every single week, there it is.

One of those questions will sting more than the rest. That’s your weakest bucket. That’s where you start. Not the bucket that sounds fun. The one that’s bleeding.

Now pick one lever inside it. The one closest to money. Not the clever one, the one where a small fix shows up in the bank.

Then comes the part most owners get backwards. They start with the tool. They see a slick AI demo and go hunting for a place to use it. That’s the tail wagging the dog.

The real question was never “which AI tool should I buy.” It’s “which lever needs pulling.” Name that first, then ask how AI pulls it harder, faster, or cheaper. The tool serves the lever, never the reverse.

And keep the first one small. Give yourself 30 days. One lever. One clear before-and-after number. One honest look at whether it moved. Speed-to-lead drops from two hours to two minutes. Repeat orders climb four points. Quote time falls from a day to an hour. Pick something you can measure by the end of the month.

The 30 days is the point. It’s long enough to show a real result and short enough that you actually finish. If it worked, you’ve got proof, and you move to the next lever with your head up. If it didn’t, you spent 30 days, not 30 thousand dollars, and you know more than you did before.

The owners who win with AI over the next two years won’t be the ones who bought the most tools or attended more workshops. They’ll be the ones who aimed.

— Sandeep Dhall. I write about how owners of real businesses make technology, AI, and growth decisions — the ones with real money on the line. Co-founder of Bridge Digital, solving real business problems with AI and custom software. Founder of SMB Capital Partners, where we buy, build, and back small businesses worth keeping.