Last Tuesday a plumbing company owner in Tampa sat on a Zoom call with me for forty minutes. He opened with: "Look, I keep hearing about AI. My competitors are talking about it. My kid showed me ChatGPT at Thanksgiving. But I have no idea what this actually does for a plumbing business."
I hear some version of that sentence every single week. Different industries, different revenue sizes, same exact gap: business owners know AI exists, know it probably matters, and have zero clue how to connect it to the daily reality of running their operation.
This post is the answer I give every one of them. Not theory. Not a pitch deck. The actual step-by-step process we use when a business owner says "how do I use AI in my business?" for the very first time.
Step 1: Forget AI for a minute. Map where your time actually goes.
The biggest mistake I see is business owners starting with the technology. They download ChatGPT, play with it for an afternoon, get a mediocre sales email, and decide AI is overhyped. That's completely backwards.
Start with the pain. Grab a pen and a blank sheet of paper. Write down every task that chews up your week that doesn't directly generate revenue. I mean the stuff that makes you mutter under your breath at 6 PM on a Wednesday.
Here's what that list usually looks like when I do this exercise with clients:
- Chasing people for information you already asked for
- Manually entering data from one system into another
- Following up on quotes, invoices, or proposals
- Building reports nobody reads but everyone demands
- Answering the same 15 customer questions over and over
- Scheduling and rescheduling appointments
McKinsey published a study in 2023 estimating that 45% of paid work activities could be automated with technology that already exists. Not future tech. Current tech. But you can't automate what you haven't mapped.
That Tampa plumber? His list was 23 items long. He'd never written them all down before. Just seeing it on paper changed his entire perspective.
Step 2: Score each item by pain and frequency
Not everything on your list is worth automating. Some tasks are annoying but only happen once a month. Others happen forty times a day and each one costs you money when it's done wrong.
Operational blind spots, AI-driven ROI breakdowns, and industry deep dives — delivered to operators who build.
I use a dead-simple scoring method with clients: rate each task on two axes. How much does it hurt when it's done late, done wrong, or forgotten entirely? And how many times per week does it happen?
Anything that scores high on both is a prime automation candidate. Anything that's high pain but low frequency might be worth a process fix instead. Anything that's low pain and low frequency — leave it alone. Automating trivial stuff is a waste of money and it poisons your team's trust in the project.
For the plumber, follow-up on open quotes scored highest. His office manager was manually tracking 40-60 outstanding quotes in a spreadsheet. She'd forget one, and a $12,000 repipe job would go to the competitor who called back first. That's not a time problem. That's a revenue leak.
Step 3: Pick ONE thing and build it right
Go small. This is where ambitious owners get themselves in trouble. They see the list, get excited, and want to automate everything in the first month. That approach fails almost every time because your team can only absorb one change at a time.
Pick the single highest-scoring item from Step 2, and build an automation that handles it completely. Not a half-measure. Not "we'll use ChatGPT to draft a follow-up." I mean the full loop: the system detects the trigger, executes the action, and confirms the result without a human touching it.
For that plumber, we built an automated quote follow-up sequence. When a quote sits untouched for 48 hours, the system texts the customer from the office number: "Hey John, just checking — did you have any questions about the estimate we sent for the water heater swap? Happy to adjust anything." If they respond, it alerts the office manager. If they ghost for another 3 days, it sends a softer second nudge. If nothing after 7 days, it marks the lead as cold and moves on.
Total build time: 4 days. The result in the first month? They recovered 3 jobs worth $28,000 combined that would have died in the spreadsheet.
A caveat I'm honest about: that $28K number is specific to a business with high-ticket jobs and a broken follow-up process. If you sell $50 products, the math is different. The principle is the same, but the dollar impact scales with your average deal size.
Step 4: Measure the win, then stack the next one
This is the part most "AI consultants" skip. You have to prove the first project worked before anyone on your team will trust you to do the second one.
Track two numbers: time saved and revenue impact. Time saved is easy — just compare how many hours your team burned on that task before versus after. Revenue impact takes more work. You need to connect the automation to actual closed deals, recovered invoices, or eliminated errors.
Once you have those numbers, share them with your team. Not in a slide deck. In a text message. "Hey team, the new follow-up system recovered 3 jobs this month. That's $28K we would have lost. Nice work." That message is worth more than any training session because it turns skeptics into advocates.
Then you go back to your list from Step 1 and pick the next highest-scoring item. Layer by layer. Each one gets easier because your team has already seen proof that automation actually works and doesn't threaten their jobs.
Harvard Business School's 2024 survey on AI adoption in mid-market companies found that businesses using a phased implementation approach had 3.1x higher long-term retention of AI systems compared to those that tried full-scale deployment. The phased approach isn't slower. It's durable.
Where most business owners get stuck
I want to be direct about two places where people stall out.
First — choosing between off-the-shelf tools and custom builds. If a tool like HubSpot or Salesforce already does what you need, don't hire someone to build it from scratch. That's an ego project, not a business decision. But if your workflow has specific logic that no off-the-shelf tool handles cleanly — weird approval chains, industry-specific compliance, multi-step processes that span four systems — you need a custom build. The honest answer is that about 60% of what we build for clients could technically be done with existing tools. The problem is that nobody on the client's team knows how to configure those tools properly, and a misconfigured HubSpot workflow is worse than no workflow at all.
Second — underestimating the people side. Your ops manager has been doing that task by hand for three years. She's fast at it. She's proud of it. When you tell her a robot is doing it now, she hears "you're not needed." You have to frame every automation as "we're giving you your time back so you can focus on [specific higher-value thing]." If you can't fill in that blank, you're automating wrong.
Still not sure where to start? Our 3-minute assessment maps your specific bottlenecks, scores them by impact, and gives you a prioritized list of what to automate first. No call, no sales pitch — just an honest diagnostic you can hand to your team on Monday morning.