Almost every founder or business owner I talk to right now wants to "go AI first." Rebuild the whole operation around automation, agents, and LLMs. It sounds like the smart move — and I get why. But I've watched it go wrong enough times that I think it's worth saying clearly: that approach usually makes things worse before it makes them better.
The trap
The thinking goes: AI is powerful, we have inefficiencies, therefore we should automate as much as possible. The problem is that when you try to automate a messy process, you don't get an efficient process — you get a fast, messy one. Automation doesn't fix problems, it amplifies whatever is already there.
I've seen small teams spend months rebuilding workflows around AI tools, only to end up with something harder to maintain, harder to explain to new people, and no more reliable than what they had before. The dream of a fully automated operation became a fragile system nobody fully understood.
What actually works
The founders who get real value from AI almost always start with one specific, small, repetitive problem. Not "how do we automate our business" — but "this one thing happens twenty times a week, takes ten minutes each time, and the steps never change." That's your first candidate.
- A form that gets filled in and then manually copied somewhere else
- A message type you always respond to the same way
- A weekly report that someone has to build from the same data every time
Start with the problem, not the technology
The right question isn't "where can we use AI?" — it's "what's costing us real time right now that follows a clear, consistent pattern?" If you can't describe the task in two sentences with no exceptions, it's probably not ready to automate.
Pick one thing. Automate just that. See if it holds up in practice. Then, once it's working quietly in the background, pick the next one. That compounding effect — small, reliable wins — is how the businesses I've worked with actually got value from AI. Not a big bang transformation, just boring problems quietly disappearing.