Everyone tells you to start using AI. Almost nobody tells you where to start, so most founders start with whatever they saw someone demo — and quietly abandon it three weeks later.
Knowing where AI creates leverage is the map. This is the first step: choosing the single task you point it at first, and proving it works before you build anything on top of it.
Find the candidates first
For one week, keep a rough log of where your hours actually go. Not a time-tracking system — a note on your phone, a line per task, five seconds each. Most founders are wrong about their own week, and the log is what corrects them.
At the end of the week you'll have a list. Now filter it down with two questions:
- Does this happen again and again? A task you do once a quarter is not worth automating, however painful it is.
- If it goes wrong, how expensive is that? A bad draft costs you five minutes. A bad invoice costs you a customer.
The beachhead test
From what survives, pick the task that eats the most hours and where being wrong is cheapest.
That combination is the whole point. Hours give you a payoff big enough to notice. Cheap mistakes give you permission to let the thing run before it is perfect — and something that has to be perfect before it runs never runs.
This almost always lands on something unglamorous. Turning the same question into the same answer for the fortieth time. Cleaning a spreadsheet. Writing the first version of a message you will edit anyway. That is the right answer, not a disappointing one.
Prove it in two weeks
Give the task a fixed trial rather than an open-ended rollout. Two weeks is enough.
Before you start, write down what "working" means in numbers you can actually check: hours saved per week, or replies sent, or a queue that stops growing. Vague goals produce vague results and no way to tell whether to continue.
Then run it on a slice, not the whole business. One channel, one customer segment, one type of document. A small blast radius is what lets you move fast without being reckless.
At the end of two weeks you will have one of three answers. It saved real time and quality held, so widen it. It saved time but quality slipped, so put a human check in front of the output and try again. Or it did neither, so drop it and go back to your list. Dropping it is a good outcome — you spent two weeks instead of two quarters finding out.
Then make it repeatable
The gap between founders who get value from AI and founders who do not is rarely the model. It is whether the thing that worked got written down.
An automation that lives in one person's head and one browser tab is a habit, not a system. Once something passes its trial, spend an hour documenting how it runs and what it touches. That hour is what turns a saved afternoon into a saved afternoon every week.
Start with one task. Prove it. Write it down. Then pick the next one off the list you already made.
Agenloo does this with founders — finding the task worth starting with, then building the agent or automation that owns it. See how we work or tell us about your business.