Book a call
All playbooks

Save time · 8 min · Free

Buy Back Your Time with AI

A 30-day way to put AI to work on one real job in your business: build a small working tool in two weeks, then spend two weeks proving it gives you time back.

Skip the plan, test something real

Most small business owners believe AI can help them, and most never get to the finish line. They stall in planning with unclear priorities, or the first step looks too big and too expensive to take.

You don't need a strategy deck to start. You need one working tool on one real job, and an honest way to tell whether it helped. That's the whole idea here: two weeks to build something small, two weeks to prove it inside the business, then a clear decision.

Pick one workflow

Start with recurring work that should be automatic: reports, follow-ups, data entry, status updates, manual handoffs. The best first candidates eat hours every week, pull data from more than one tool, and are often late because someone has to do them by hand.

Keep it practical and business-facing. A good first test saves time, removes friction, or makes a job run more smoothly. It usually looks like an automation, a small internal tool, or a better version of a workflow you already run, not a research project.

You also need someone who can explain the job step by step: what comes in, what they do with it, and what good output looks like. If nobody can describe the workflow, AI can't follow it either. Pick a different one.

Write down the baseline

Before you build anything, write down what the job costs today: hours per week, who does it, how often it's late, and how often it has mistakes. Ten minutes with the person who does the work is usually enough.

Without a baseline, you'll finish the test with a feeling instead of an answer. The Automation ROI Calculator can help you put a yearly number on the hours.

Weeks 1–2: build the smallest version that does real work

Build something the person who does the job can actually use on real inputs, not a demo on sample data. Use the simplest mix that works: an AI assistant with clear written instructions and examples of good output, a connection to the tools where the data lives, and a small internal tool only if the job needs one.

Keep a person reviewing the output at first. The goal of the first version is to take the repetitive part off someone's plate, not to remove their judgment. Cut anything that isn't needed for the core job; edge cases can wait until you know the tool earns its place.

Weeks 3–4: prove it inside the business

Use the tool on the real work for two weeks. Track the same things you wrote down in the baseline: hours spent, how often it's late, how often it's wrong, and what it would have cost to keep doing it by hand.

Keep a short log of where it broke or needed a fix. That list tells you what a production version needs, or whether the job was a poor fit for AI in the first place.

Decide what's next

At the end of 30 days there are three honest outcomes. If it clearly saves time, harden it for everyday use: make it reliable, add monitoring so you know when it fails, and decide who maintains it. If it almost works, adjust the scope and run another two weeks. If it doesn't help, stop and pick the next workflow. A test that didn't work still tells you something useful, and it only took a month.

Once one workflow is running, repeat the process. Add new workflows one at a time, measure each against its own baseline, and keep improving the ones you already run as your tools and work change.

Checklist

  • List your recurring manual jobs and rank them by hours per week.
  • Pick one that pulls from more than one tool and that someone can explain step by step.
  • Write down the baseline: hours, owner, how often it's late, how often it's wrong.
  • Build the smallest working version in two weeks and use it on real work.
  • Track the same numbers for two more weeks, then keep it, adjust it, or drop it.

Want help running the test?

Run it yourself, bring it to a 90-minute, $199 working session, or have me build it.