Almost every business owner I talk to has the same instinct right now: "we should be doing something with AI." Very few can tell me what that something is. So they end up with a scattered pile of half-tried tools — a chatbot nobody finished setting up, a subscription to something with "AI" in the name, a pilot that quietly died after the person who started it got busy.
That's not a technology problem. It's a prioritization problem. The businesses actually getting value from AI right now aren't the ones with the most tools — they're the ones that picked the right place to start.
Here's the three-question framework I use with clients to find that starting point.
1. Where is the work repetitive, but not simple?
Skip the truly simple, one-click tasks — automating those barely moves the needle. Look instead for work that's repetitive and requires judgment: writing similar-but-different client emails, summarizing calls, triaging incoming requests, drafting first versions of reports. That's the sweet spot where AI genuinely saves hours instead of just shuffling a button.
2. Where does a delay actually cost you something?
Not all slow processes matter equally. A slow monthly report is an annoyance. A slow response to a hot lead is a lost sale. Rank your candidate workflows by what a delay actually costs — in lost revenue, lost customers, or lost time from your most expensive people — and let that ranking decide the order, not how "exciting" the AI use case sounds.
3. Can you tell, within two weeks, if it worked?
Your first AI project should be small enough to prove itself fast. If success can only be judged after a quarter, you've picked a strategy initiative, not a starting project. Pick something where you'll know within two weeks whether it's actually saving time or producing better output — then use that early win to justify the next one.
"The goal of your first AI project isn't to be impressive. It's to be finished, used, and clearly worth repeating."
Putting it together
Run every workflow you're considering through all three filters: repetitive-but-judgment-heavy, meaningfully costly when slow, and fast to prove out. Usually only one or two workflows survive all three — and those are the ones worth automating first.
If you want a second set of eyes on this for your own business, that's exactly what a free intro consult is for — no pitch, just an honest read on where to start.