How One Business Operator Saved 3,500+ Hours Using AI, and What That Means for Your Operation

I was running a marketing consultancy in West Michigan with a small team of contractors. The contractors were handling client work. I was the only one in the operation using AI to go faster.

That context matters. This isn't a story about a company that added AI to an existing tech team. It's about one person who figured out, project by project, that the gap between having an idea and having a working thing had basically closed.

The breakdown spans five categories: full-stack web applications, smaller tools and utilities, content and sales systems, research tasks, and custom AI tools built for recurring work.

The web applications are the most concrete place to start. Over three years I shipped eight production-grade tools with AI assistance. A full SaaS product for email signatures with Stripe payments, OAuth, a drag-and-drop interface, and team features. An internal audit tool I use with clients. A hiring assessment app that covers seven personality frameworks, generates PDFs, and handles its own email delivery. A blog publishing pipeline that runs on a monthly schedule with about 17 minutes of my input.

A developer would have quoted me $40,000 to $100,000 for any one of those. I'm not a developer, and I built all of them.

The smaller tools are nine working web apps for specific needs: service page generators, client dashboards, calculation tools. Things that would have cost $5,000 to $20,000 apiece and taken months to schedule with a contractor. Each one took a few hours of my time to specify and ship.

The content and sales systems include a full business book built around a proprietary hiring framework, five complete speaking presentations with scripts, a sales process with intake questionnaires and discovery call scripts, and the autonomous blog I mentioned. A ghostwriter charges $30,000 to $80,000 for a business book. This one exists because AI made it possible to work on it in the margins of a normal work week.

Research tasks: 35 discrete projects that used to take full days of manual searching, compressed into hours. Competitor analysis, market sizing, lead generation strategy, content planning.

Custom AI tools: nine purpose-built tools I use daily for drafting, headline generation, client content, and a few things I built just because I could.

I can't read code. No formal training in software development. Before AI, if I wanted to build something, my options were: hire a contractor, find a no-code workaround, or skip it.

Most of what's in this list wouldn't exist without AI. I wouldn't have hired a developer to build a calculation widget for client presentations. The cost wouldn't have been worth it. With AI, the threshold for “worth building” dropped to almost nothing. So I built more things, and some of those things turned out to be genuinely useful.

For years the constraint was the cost of execution. Hiring a developer, finding a workaround, or dropping the idea. AI removed that constraint, and once it goes near zero, a lot of things that were never worth building before suddenly are.

My work is heavy on knowledge work: writing, building, research, systems design. A plumbing company has a different mix, but the math works the same way.

Where service businesses lose the most hours per week is not glamorous. The follow-up email on an estimate that went out two weeks ago and hasn't been touched since, because the person who sent it was also running jobs. After-call documentation reconstructed from memory at 5 PM, usually incomplete. Appointment reminders that live in someone's head until the customer calls to reschedule.

An HVAC company with four technicians is probably losing 10 to 20 hours per week across those three categories. That's a workflow gap, and it's the kind of gap AI handles well.

A follow-up sequence for unsigned estimates takes about 3 to 5 hours to build in Make.com or Zapier. Once it's running, it runs every time, without anyone managing it. Closed estimate data from several HVAC companies shows 20% to 35% of unsigned estimates close on the second or third follow-up. If you're sending 40 estimates a month and not following up at all, the setup time pays back fast.

The same pattern applies to review requests, appointment confirmations, service summaries, and dispatch documentation. None of it is complicated to build. All of it requires someone to actually set it up, which is the part most service businesses skip.

I built tools that nobody used, including me. A client dashboard that turned out to be more work to populate than it saved. A wireframe builder I convinced myself I'd use weekly before realizing I don't make wireframes that often. A couple of content tools that were outpaced by the next version of the AI they were built around.

I also got work wrong in ways I didn't expect. Research tasks where the AI kept generating citations I couldn't verify. Writing tasks where my edits took longer than starting from scratch. One tool I rebuilt three times before the structure was actually right.

You ship faster and you fail faster. The wins compound and the misses still happen. The net is overwhelmingly positive, but it's not a clean story, and anyone telling you it is has either done very little of it or is leaving something out.

If you're a service business owner reading this and wondering whether any of it applies to you, some of it does and it's probably not the most impressive parts.

You're not building SaaS products. You're automating follow-up sequences and dispatch documentation and review requests. The tools are simpler. The setup is shorter. The payback is measured in weeks, not months.

The question is which ones, in your specific operation, in the order that makes sense for your workflow and your team. That's the question the AI Audit answers. I go through your operation the same way I went through mine, and come back with a specific list of what to build and where to start. Not a framework or a deck full of general AI principles. A roadmap.

If you want to see what that looks like before you commit to it, the discovery call is free and there's no pitch at the end.