This is the third post in a series working through one industry at a time. This week it’s field service businesses: think HVAC, plumbing, electrical, and property maintenance.

While I haven’t run a trades business, what I have done is build the operations software underneath businesses that live and die by their schedule (for instance, ResortSteward), and I build integration layers that wire AI systems into a company’s calendar and records. That’s the technology underneath the solutions I’m going to cover for a field service business.

My standard caveat applies. Neither I (nor anyone else!) can tell you from the outside which of these fits your shop or in what order you should implement them. What I can do is point at where my experience says AI earns its keep in a business that runs on trucks.

The Work That Eats the Week

This is the first industry in the series where the constraint is physical. A steel service centre is limited by what it can process and a professional firm by billable attention, but what you sell is technician hours in a vehicle. Almost everything worth pointing AI at here is aimed at making those hours productive and billable.

There are four time thieves in a field service business: drive time, the second trip (for instance, because the part wasn’t on the truck), the idle slot that opens when a customer cancels at the last minute, and the paperwork done late at night or not at all.

What I’ve Already Covered

If you’ve been following this series, you’ve already seen some of the solutions. Your phone is the front door of a field service business and usually the first place to point AI, and I covered that in “When AI Picks Up the Phone”. Retrieval over internal documents like policies and warranty terms is the pattern from “When AI Reads Your Documents”. And a quoting tool that encodes your own flat rate book and creates quotes for a person to send is the solution I highlighted for steel RFQs and professional services proposals. Those solutions are directly applicable to a field services business, so it’s worth reading those posts if you haven’t already.

However, there is one version of AI-assisted information retrieval that is different enough to separate out. A technician standing in front of a misbehaving unit doesn’t run a keyword search, they describe a symptom in ordinary words. The useful system reads the manufacturer manuals (the authority), alongside your own closed tickets (the experience: we’ve seen this before on this model and here’s what fixed it). Then it points at the manual and the section with a link rather than paraphrasing it, because someone working on a live system needs the real document. That’s close to a perfect fit for what Docora does, so I’ll point you there. This kind of solution gives inexperienced employees the equivalent of a senior technician at their side.

Use Cases

The schedule that re-plans itself. Many shops build the next day’s schedule by hand…and then rebuild it four times before lunch as jobs run long, customers cancel, emergencies arise, or employees call in sick. That can result in extra travel time, idle slots, or dissatisfied customers. What helps is scheduling that understands capacity, skills, and territory and proposes a new plan when the day changes: the right technician, properly certified, closest to the job, with the right parts aboard their truck.

The second trip. An expensive failure in this business is arriving without the right part. An intelligent AI-assisted system can help mitigate this even before dispatch — the job description, the equipment on file, the service history, and past jobs that looked like this one can be read together to predict what the technician will need and to flag the diagnostic questions that should be asked on intake calls.

The write-up from the driveway, and the job you noticed while you were there. The paperwork half of this solution is quite straightforward: the technician talks or photographs, and the system drafts the work order, the invoice lines, and the customer summary. But the more interesting half is what else the technician saw while at the job site. Although they were there to fix one thing, they likely noticed other potential issues (e.g., the water heater starting to corrode, the panel with no capacity left, the unit that won’t survive another winter). There’s no better lead for you than your own qualified professional physically looking at a specific problem in a specific building. In many cases this opportunity is lost because there’s nowhere structured to put it, and the technician is already late for the next call anyways. AI can help by supporting a process that turns those observations into a record, separating safety concerns from deferred maintenance from straightforward upgrades, then drafting a follow-up at the right time using the technician’s own observations.

The customers you already have. Every one of those observations becomes another line in the record of what you’ve installed and seen at each customer’s site. Many shops treat that as service history, but in reality it’s a replacement pipeline where no prospect list must be developed (or bought!) and nobody is contacted cold.

So why does this need AI rather than a report and a mail merge? Realistically, historical records are often messy. That isn’t a criticism of anyone, it’s just the normal condition of a business that’s been running twenty years. A query needs consistent fields and structured data, but historical data isn’t usually structured this way. Most shops aren’t actively mining historical records because it’s extremely time-consuming, but pulling reliable structure out of inconsistent, human-written history is one of the things AI is really good at.

Once an AI-assisted system has pulled in historical data, it can flag equipment near the end of its life or that is beginning to present problems. It can then develop outreach materials that have the reasoning included — specifics like “this unit is twelve years old, we’ve been out three times over the last three years, and twice for the same fault, so you’d be better off upgrading to this.” It can check which rebates and efficiency programs actually apply to each customer and include that as part of the rationale. Then it drafts the conversation for someone to review and send.

Where AI Is the Wrong Tool Here

AI should be deployed in dispatch and routing as a decision support tool, not as an autonomous tool that fully controls where and when your people go. While the AI can make fully formed dispatch plans, often there’s still an element of judgement in deciding how your employees will be dispatched.

An important area where judgement cannot be delegated is safety. Technician safety observations are not sales opportunities to be followed up on through the sales pipeline. Something like a gas concern, a scorched panel, or a venting problem needs to go to a person immediately and gets handled as an urgent call. The separation of sales opportunities and safety concerns must be deliberate, because a naïve approach that simply sorts by potential revenue will happily file a safety issue behind a more profitable upgrade.

Finally, a poorly designed upgrade recommender will also always find upgrades, including in cases where it isn’t warranted. In a trade that runs on referrals, that’s a reputational risk. Recommending an unneeded upgrade or getting an incentive wrong is worse than never mentioning it at all, and the system must be designed with guardrails that prevent this.

And then there’s the thing I say for every industry. If your schedule is chaotic because the process underneath it is broken, or your records don’t match what’s actually installed, AI is just going to misroute trucks faster and recommend replacing things that were never there. You must fix the process and the records first.

Where to Start

I can’t tell you from the outside which of these fits your shop, and the honest answer usually depends on what your dispatch board, your second-trip rate, and sales into past customers actually look like. That’s what a Discovery Day is for: one day, in your business, working out the one or two places this would earn its keep and saying plainly where it wouldn’t.