# Where to Start With AI When You're Too Busy to Figure It Out > A practical first step for HVAC, dental, and law shops who tried ChatGPT and got nowhere. Not a list of 50 tools, but a decision framework that starts with your actual work. - **Author:** Cal - **Published:** 2026-05-31T00:00:00.000Z - **Reading time:** 9 min - **Pillar:** Getting Started - **Canonical:** https://www.ownersmethod.com/blog/where-to-start-with-ai-small-business --- You've probably tried ChatGPT. Maybe you watched a few demos. Got a generic response about your business and thought, okay, so what? Nothing changed on Monday morning. That's not your fault. The problem isn't that you're not technical enough. The problem is that every article about "AI for small business" is written for some abstract business that doesn't exist. Not for the HVAC company with six techs and an office manager fielding 40 calls a week. Not for the dental practice with three chairs and a front desk that handles scheduling, billing, and the phone simultaneously. This article is for those businesses. If you run a small business and you've been stuck on where to start with AI, I'm going to tell you exactly that. Not with a list of 50 tools, but with a decision framework that starts with your actual work. **The short answer:** Start with the single task that costs you the most time and happens more than twice a week, not with a tool, and not with five tools at once. Pick that one task, find the simplest tool that handles it, and give it four weeks before you add anything else. Everything else in this guide follows from that one decision.

"How can AI help my business?" is the wrong first question. The right one is: what am I doing by hand that a tool could do faster?

--- ## You've probably already tried (and gotten nowhere) Here's what usually happens. An owner, let's say a plumber with 12 employees, hears about AI. Tries ChatGPT. Types something like "how can AI help my plumbing business?" Gets back a generic five-paragraph answer that could apply to any business on earth. Closes the tab. Or they try to use it for a specific thing, writing a job posting, maybe, and the first draft is fine but still needs an hour of editing. So they wonder: is this even faster? The problem is the question they started with. "How can AI help my business?" is the wrong first question. It's like walking into a hardware store and asking "what tool should I buy?" You need to know what you're building first. The right first question is: **What task am I doing by hand that a tool could do faster?** That's it. That single question, asked honestly about your own workday, will get you further in an afternoon than six months of reading AI newsletters. --- ## Where to start with AI: the real question isn't "what tools should I use?" The translation gap in AI isn't complexity. The tools are genuinely easy to use. The gap is this: nobody has translated "AI can do this" into "here is exactly how to set it up for _your_ business." Before I started The Owner's Method, I spent years building the operations and technology stack at Fields Residential, a company that did over $100 million in new-construction homes with a team of five people. The way we stayed lean was by asking one question about every workflow: _Is a person doing this by hand when a tool could do it faster?_ Investor fundraising used to take days of calls and paperwork. We got it to a one-click email. Contract generation went from several hours to less than 10 minutes. We didn't do this because we were technical. We did it because we got disciplined about asking the question. That's the owner's eye. And it's the same thing AI requires. You don't need to understand how a large language model works. You need to understand your own workflows well enough to identify which ones are repetitive, rule-based, and eating time that could be spent elsewhere. --- ## Start with the task that costs you the most time Write down every task you do, or your team does, more than twice a week. Not "running the business." Specific tasks. Things like: - Answering the same five questions customers ask via phone or email - Following up on unpaid invoices or estimates that haven't been accepted - Typing up meeting notes or call summaries - Scheduling, rescheduling, and confirming appointments - Writing job descriptions, performance reviews, or email replies - Responding to Google reviews Now circle the one that takes the most time in aggregate across your whole week. That's your starting point. Here's what this looks like in practice for three types of businesses I work with. **HVAC company:** The front office spends 3-4 hours a day on inbound calls, most of them for scheduling, basic diagnostic questions ("why is my unit making that sound?"), and quote follow-ups. In the shops I've worked with, an AI receptionist (Housecall Pro and several others now offer one) takes a real chunk of that routine call volume without a human on the line, it books the job, captures the contact, logs the call. The office manager still handles the complex ones, but the routine volume drops. **Dental practice:** The front desk handles appointment reminders manually, calls or texts sent one by one, each requiring a staff member to stop what they're doing. Most modern dental software (Dentrix, Eaglesoft, Curve) has automated reminder features that patients already expect. If yours doesn't have them on, that's 2-3 hours a week back for free. Not AI exactly, but same principle: stop doing by hand what a tool already handles. **Law firm (small practice):** Two attorneys and a paralegal. Every client call used to require someone to take notes, then type up a summary afterward. They started using Fathom, a free AI meeting transcription tool, on every call. At the end of the call, Fathom gives them an accurate transcript and a summary with action items. In a practice like that, the note-writing was running 30-45 minutes per call; across the ten or so client calls in a week, that's roughly 5-6 hours back for the team. (These are the kinds of figures I see in practice, not survey data.)
5-6 hrs

a week back for that small law practice, across ten client calls, from one free tool aimed at one repetitive task.

The pattern in all three: they didn't start with "AI strategy." They started with a specific task that was eating time, found the simplest tool that addressed it, and [recovered five or more hours a week](/blog/how-to-save-5-hours-a-week-ai) in the process. --- ## The three tools worth setting up first (for most service businesses) I am not going to give you a list of 50 tools. That's the problem, not the solution. For most local service businesses, trades or professional services, the first tools worth your time are in three categories. Here's what they are and why. | Tool | What it's for | Cost | Best for | | -------------------------------------------------------------------------- | ------------------------------------------------------------------ | -------------- | ----------------------------------------------------- | | Fathom | Call & meeting transcription with summaries | Free tier | Anyone with regular client, team, or vendor calls | | ChatGPT or Claude | First drafts of pattern writing (emails, postings, review replies) | $0-20/mo | Owners who write the same kinds of things repeatedly | | Your existing software's AI features (Jobber, Housecall Pro, ServiceTitan) | Reminders, AI receptionist, smart scheduling | Already paying | Trades businesses already on a field-service platform | ### 1. AI call and meeting transcription (Fathom) **What it does:** Records your calls and meetings, transcribes them accurately, and gives you a summary with action items. Automatically. **Why this one first:** Almost every service business has calls and meetings where someone is supposed to be taking notes and half the time isn't. Fathom makes this a non-issue. The free tier is genuinely functional. You set it up once, add it to your calendar, and it joins every video call automatically. **Who it helps most:** Anybody who has client calls, team meetings, vendor calls, or internal check-ins. That's most businesses. An HVAC company doing site assessments over video call, a dentist doing a patient consultation, a bookkeeper reviewing a client's P&L, all the same problem, same tool. **Time back:** In practice, it saves roughly 30-40 minutes per meeting that previously required someone to take notes and write them up. At five meetings a week, that's 2-3 hours. ### 2. ChatGPT or Claude with a specific prompt (for writing) **What it does:** Writes first drafts of anything that follows a pattern: job postings, estimate follow-up emails, review responses, staff announcements, service descriptions. **Why this one second:** Writing is where most owners underestimate the time they spend. A good job posting takes 45-60 minutes to write from scratch. A response to a negative Google review can take 20 minutes if you're trying to get the tone right. ChatGPT can do a credible first draft in 30 seconds. You edit it, that takes 10 minutes. You just got 35-45 minutes back on that one task. **What actually works:** Specific prompts beat vague ones by a mile. "Write a response to this Google review" and then pasting the review gets you something usable. "How can I improve my marketing?" gets you a lecture. Here's the prompt format that works: _"I run a [type of business] in [city]. Write a [document type] that [specific goal]. Tone should be [professional/friendly/direct]. Here's the context: [paste the relevant information]."_ That's it. Specific business type, specific output, specific tone, specific context. You'll get a draft that's most of the way there in under a minute. ### 3. Your existing software's AI features (you probably already have them) **What it does:** The field service software you're already paying for, Jobber, Housecall Pro, ServiceTitan, or their equivalents, has been adding AI features over the last two years. Automated follow-up reminders. AI receptionist for after-hours calls. Smart scheduling. Estimate follow-up sequences. **Why this one is often the fastest win:** You're already paying for the platform. The features are already there. Most owners have never turned them on. Before you spend a single dollar on a new tool, log into whatever scheduling or CRM software you use and look at the features you haven't enabled. For most trades businesses, there are 2-3 hours of saved time sitting there in the settings menu. --- ## What doesn't work (and why most owners give up) I want to be honest here, because most AI content glosses over this part. **Vague prompts.** "Help me grow my business" will get you a useless generic response. "Write a follow-up email to a customer who got a roofing estimate from me three weeks ago and hasn't responded, keep it brief, friendly, and ask if they have questions" will get you something you can actually send. **Starting with five tools at once.** This is how AI projects fail. You set up five different tools over a weekend, none of them are integrated into your actual workflow yet, and three weeks later you've stopped using four of them. Pick one task. Get one tool working for that task. Use it for four weeks. Then add the next. **Expecting AI to run the business.** AI does not know your customers. It does not have judgment about which tech is right for the job. It does not understand the relationship between you and a longtime client. It handles pattern and repetition well. Judgment and relationship, not at all. This is also why AI rarely replaces whole roles: according to [McKinsey's State of AI survey](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai) (2025), only about 17% of organizations saw any AI-driven headcount reduction in the past year, most redeployed the freed-up time instead of cutting people. **The demo problem.** AI demos always show the best case. The vendor shows you a chatbot handling a perfect customer conversation. In real life, your customers will ask things the chatbot doesn't expect, and then it either fails or gives them the wrong answer. Any AI tool that touches customers needs a clear handoff to a human when it hits its limits. Build that handoff before you launch it. **Chasing the newest thing.** A new AI tool comes out every week. Most of them aren't meaningfully better than what you already have. The one who gets value out of AI isn't the one who adopts every new tool. It's the one who uses one tool deeply enough that it actually changes their workflow. ([Whether AI is worth it at all for your business](/blog/is-ai-worth-it-for-small-business) is worth settling before that first investment.) I'll tell you what I've seen actually happen across dozens of businesses: the ones who got the most back spent two weeks doing one thing well, not two weeks installing everything and doing nothing well. --- ## What to do if you have 10+ workflows that are a mess Here's the honest version. If you can identify two or three specific tasks that each eat two or more hours a week, you can probably start yourself. Use this article. Pick the task, pick the tool, give it four weeks. If you have ten or more workflows that feel like they've gotten away from you (the phone is ringing too much, the follow-up is falling through cracks, the team is doing the same work twice, and you don't know where to start), someone looking at your actual situation will save you six months of trial and error. Not a tool vendor. Not another listicle. Someone who sits down, looks at your real workflows, and tells you the three things worth changing first. That's what the [45-Minute AI Assessment](/schedule-your-ai-assessment) is. You get a recorded call, a curated shortlist of 3-7 specific moves worth making for your business, and a four-day plan to get the first one running. Not generic advice. Your business specifically. If you're not sure whether you need that yet, the [free AI Scorecard quiz](/free-ai-scorecard) places you on a five-stage operational maturity scale and tells you what's realistic to expect at your current stage. Takes two minutes. --- ## One thing to do this week The biggest enemy here is overwhelm. You read something like this and think "I need to audit my whole business" and then don't do anything. So here's the one thing. Not five things. One. **This week: write down every task you or your team repeats more than twice a week.** That's it. No tools. No AI. Just a list. Takes 20 minutes. Do it in a notebook, in a Google Doc, in a text message to yourself, doesn't matter. Once you have that list, you'll know where to start. The highest-time, highest-frequency item on that list is your first AI project. Everything in this article will make more sense once you're looking at your specific list. AI works. The instructions have been missing. This is where they start. --- _Related reading:_ - [How to Save 5+ Hours a Week as a Service Business Owner (Using AI)](/blog/how-to-save-5-hours-a-week-ai) - [Is AI Worth It for a Small Business? An Honest Answer With Real Numbers](/blog/is-ai-worth-it-for-small-business) - [AI for HVAC Companies: The No-Nonsense Guide for Owners](/blog/ai-for-hvac-companies) - [AI for Dental Practices: The Admin Hours You Can Actually Get Back](/blog/ai-for-dental-practice) - [Take the free AI Scorecard, find out where you actually stand (2 minutes)](/free-ai-scorecard) _External references:_ - [McKinsey, The State of AI 2025 (workforce impact: ~17% saw AI-driven headcount reductions)](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai) - [U.S. Chamber of Commerce, AI Training Guide for Small Business](https://www.uschamber.com/co/start/strategy/ai-training-guide-small-business)