Fixing the Broken AI Strategy in Field Service Businesses

By now you've seen the AI news, it's all over the place. Every tech company is adding AI to their offerings and Wall Street is bullish on anything with AI attached to it. This isn't just in the United States, it's the same story everywhere. Business executives are struggling to find the right use cases where AI can actually create efficiency, reduce costs, or bring in additional revenue.

What the headlines don't tell you is that the companies successfully implementing AI are not cutting people, they're hiring more. The ones doing layoffs are actually the companies still trying to figure AI out. Their executives want to show investors a quick win through cost reduction and better revenue numbers, but that approach doesn't hold up over time.

The companies that are going to come out ahead are the ones that sat down first and took a hard look at their people, their processes, their workflows, and their data before doing anything else.

To put this in perspective, research from PlanGrid and FMI found that service and construction professionals already spend about 35% of their time on non-productive activities, things like hunting down information, resolving conflicts, and dealing with rework. That's over 14 hours a week per person that isn't moving the business forward. AI, when applied correctly, is the single best tool available to claw that time back.

And yet most businesses haven't even started. A 2026 ServiceTitan report covering over 1,000 service contractors found that 79% aren't using AI at all, and only 4% are using AI features built into their existing software. The ones who have taken the leap are already seeing results, with 48% of early adopters reporting a meaningful jump in productivity.

So where do most go wrong when they do try? Take this scenario. You've set up AI to qualify inbound inquiries. A call comes in from a customer who is already unhappy and wants to speak to someone about an escalation. The AI takes it. Their frustration gets worse because the AI cannot pick up on their tone or understand the emotional weight of the situation. You've now lost that customer. And in a service business, that customer doesn't just leave quietly. They leave a Google review or post in a local Facebook group. Word of mouth is the lifeblood of a service business and one bad automated interaction can end up costing far more than the salary of the person it was meant to replace.

This is where a simple distinction helps. There are transactional tasks and there are relational tasks. Scheduling, invoicing, dispatching, follow-up reminders, booking confirmations, these are transactional. A customer calling about a complaint, a client asking why your quote is higher than a competitor, someone wanting to know why their invoice doesn't look right, those are relational. AI belongs in the first category and humans belong in the second. If you use that as your filter before deciding where to deploy AI, you'll avoid most of the costly mistakes.

When you're starting out, pick a workflow or manual process that doesn't involve customers directly. A good starting point for most service businesses is something like automated appointment reminders, follow-up messages after a job is completed, or generating the first draft of an invoice based on the work order. These are tasks your team does dozens of times a week, they follow a predictable pattern, and a mistake in any of them doesn't damage the customer relationship the way a bad phone call would. Your office admin team handles a lot of this kind of repetitive work every day. Automate those tasks and that same team can now spend more time on inbound calls or outbound business development, work that actually grows revenue.

Once that pilot runs without issues, you can move to something like automating your website chat to book appointments around the clock. But you have to set clear rules. If a customer is frustrated or unhappy, the system should book them with a human immediately rather than asking more questions.

Before you expand though, it's worth being honest about whether your pilot is actually ready. A good sign is when the AI is handling the task consistently without anyone needing to manually fix the outputs, customers aren't noticing anything different, and your team isn't spending time cleaning up after it. If two out of three of those are true you're probably ready to take the next step. If your team is still regularly correcting what the AI produces, that process needs more work before you move on.

Because at the end of the day people want to deal with people. AI can assist with research, planning, and the behind-the-scenes tasks, but when there's a real customer interaction involved, you need a human. That's not a workaround, that's the right design.

The work that used to take five people can now be done by one or two, and that's actually a good thing if you think about it correctly. Those people can now focus on quality over quantity.

One more thing worth knowing. AI still hallucinates, still cites outdated sources, still produces numbers you can't always trust. That's why having a human review the details isn't just a nice-to-have, it's the whole point of human-in-the-loop.

If you have two or three office admin staff managing dispatch, sales, and general admin, the advice here is straightforward. Don't let them go. Ask them to do a better job at what only humans can do well, which is connecting with customers. Have them watch how AI is handling the calls and chats, identify what's working and what needs fixing, and check whether the generated invoices are actually accurate. That feedback makes the system better over time, and that's something AI cannot do on its own.