Is your service business ready for AI?

Most AI frustration in a service business starts before the tool is ever switched on. If you aren't ready for AI and don't know it, the setup goes badly, the results are annoying, and you walk away thinking AI doesn't work for a business like yours. Usually the tool was fine.

Four questions tell you whether you're ready. Can the AI read what you know? Who checks its work? Which task would you celebrate never doing again? And will a customer see what it makes?

AI multiplies whatever you give it

People who build AI tools call it a force multiplier. Point it at a clear problem with good information behind it and one person can get a lot more done. Point it at a fuzzy problem, with the information split between your head, your truck, and three apps, and it goes wrong just as fast.

So start with a problem that's reasonably well defined, and keep it small while you learn what AI can and can't do.

You're also not the first person to ask whether you're ready. Readiness checklists already exist, and Indiana has its own: TechPoint's AI Adoption Playbook for small and mid-sized businesses, and a short readiness assessment from the state's IN AI program. Both are written with an office and an IT budget in mind. The four questions below are our shorter version for a business that runs out of a truck.

1. Can the AI read what you know?

AI works best with structured information it can reach: a database, a spreadsheet, a document, or software that has an AI connector. It can't do anything with what's in your head.

If your prices, your service area, and how you handle a reschedule all live in your memory, step one is getting them out and onto something the AI can read. A Google Doc is a great place to start. Write down how you price a job, what's included, and what you won't do. That doc pays off even if you never use AI, because it's the same thing you'd hand a new hire.

A word on connectors. Plenty of software has one now. Crowfly does, and a lot of other good software does something very similar. Don't let a connector be the reason you pick a product or stay stuck with one. Pick what fits how you work, then check that it can talk to the AI tools you use, and how much it lets them do. Some connectors cover a handful of actions. We built Crowfly so that anything a person can do in it, an AI agent can do too, and anything the agent can do, you can do yourself.

2. Who checks the work?

Don't trust AI with anything until it has earned it, the same as a new employee. You wouldn't let someone in their first week send quotes without looking them over.

So before you start, decide where the human review step is. Who reads the output before it goes anywhere, and what happens when it's wrong? A wrong price in a draft you read costs you a minute. A wrong price the AI already sent to a customer costs you the job or the margin.

Owners who stuck with AI describe the same change. One small business owner said it went "from scary and useless to a key part of my day when I stopped using it as a source of truth, and started using it as an assistant" (source).

The checking takes time, and the skeptics are right that it can eat whatever you saved. That's why the next question matters so much.

3. Which task would you celebrate never doing again?

Picture this. You wake up tomorrow and someone tells you that you never have to do a certain task again. Would you cheer? Cry happy tears because you'll never line up the columns on a quote again? Or would you say "oh, that's cool, I guess"? That task can stay on your plate.

Aim AI at the one that makes you cheer. It should eat real hours, you should dislike it, and it should be small enough that checking the output doesn't take as long as doing it.

Be honest about one thing here. Some of the first tasks people name, like appointment reminders and booking confirmations, don't need AI at all. Scheduling software has been sending those for over a decade. If a plain automation does the job, use it. And if something already works fine, automating it only adds moving parts.

4. Will a customer see it?

This is where owners get burned.

A lot of people are turned off by anything that's obviously AI-generated. Once you start using AI yourself, you'll notice how many of the emails in your inbox it wrote, and it'll bother you too. So don't blame your customers for noticing, and don't write it off as older customers who won't talk to a machine. Plenty of Gen Z use ChatGPT every day and still can't stand AI-generated filler. The dislike shows up everywhere.

One business owner said they delete and blacklist any company that sends them AI-written proposals or emails (source). If you go all in on AI and everyone can tell, it loses you customers.

Anything a customer will see should get a person's eyes on it before it goes out.

The AI receptionist question

Big companies are rolling out AI receptionists, and vendors are pitching them hard to small shops. Put yourself in your customer's shoes. Would you rather talk to an AI, or leave a message and actually get a call back?

The answer depends on your market. You'll find plenty of advice saying people hate AI receptionists and hang up. Maybe your customers do, maybe they don't. A lot of people do prefer talking to a person.

So measure before you buy. How many calls are you actually missing, and are you calling those people back? One person advising a 32-person shop said to count missed calls for 10 days before buying anything. If the number is small, or the real problem is slow callbacks, there are cheaper fixes than an AI receptionist that runs hundreds of dollars a month (that's a lot of missed calls to break even on). A voicemail greeting that promises a same-day call back works. So does letting customers text the same number with their name, the problem, and a photo.

Or take a share of those calls off the phone entirely. With Crowfly, customers can book from your booking page, picking a time that fits your route and attaching notes and photos of the job. For work that needs your eyes before it goes on the calendar, they fill out an intake request instead. Either way, you get a text the moment it lands, so you know what you're walking into before you call anyone back. Then you write the quote yourself and send it from the same place. Nothing on the customer's end is a bot.

Before you spend anything

Pick the one problem that takes the most of your time. Then try the intern test: could you describe it well enough for a pretty good intern who knows nothing about your business? What goes in, what comes out, what a good result looks like, and when to stop and ask you. If you can't write that down, the AI can't do it either. Open a Google Doc and try.

Then decide what solving it is worth. Put a dollar figure on the hours it eats each month. That's your ceiling for any tool, subscription, or setup fee.

Consultants

Hiring someone to set it up is the easy on-ramp, and there are a lot of people happy to do it. Some are very good. A good one sets up something that works well the first time, without you having to learn the plumbing behind it. Some aren't, and they'll sell you a big system for a problem you haven't defined. As one small business commenter put it, "If you don't have a specific problem you don't need a consultant" (source). Show up with your intern doc and your dollar figure and you'll be able to tell the two apart quickly.

Building it yourself

Once AI is writing code for you, it's tempting to build your own version of the app you pay for and drop the subscription. You could. But software engineering involves a lot that isn't obvious from the outside: security, backups, updates, what happens when something breaks at 7am on a Saturday. The subscription pays for all of that. For most service businesses, your hours are worth more on jobs than on maintaining homemade software.

Start with the doc. It takes one evening and tells you whether you're ready.