AI for the trades

Where AI helps a service business, and where it doesn't

AI tools are good at a specific kind of work: reading, summarizing, drafting, and pulling information out of messy inputs. They are not good at knowing things they were never told, making judgment calls about your customers, or standing behind a result. Keeping that split in mind makes it much easier to decide where to use them.

Good fits are tasks where a draft saves time and a person can check the result quickly. Turning a rough note into a clear paragraph for a proposal. Reading a model and serial number off a photo of a nameplate. Answering a question like which jobs are scheduled for a customer next week. Suggesting checklist items for a type of job you have not written a checklist for yet.

Poor fits are tasks where a wrong answer is costly and hard to spot. Diagnosing equipment safety issues. Deciding what to charge. Making promises to a customer. Anything involving code compliance or a sign-off that carries your license. AI can help you prepare for those decisions, but the decision stays with a qualified person.

Also be wary of tasks you do not understand yourself. If you cannot tell whether the output is right, you cannot use it safely. AI works best as an assistant to someone who already knows the job.

Start with one small, low-risk task that eats time every week. Try it for a few weeks, check the results, and decide based on what you see rather than what a vendor says.

In Servora, the AI assistant can read customers, assets, and the schedule to answer questions, suggest checklist items, and offer dispatch suggestions on the schedule.

• Good fit: drafting, summarizing, reading information off photos and documents, answering questions about your own records.
• Poor fit: pricing decisions, safety and compliance judgments, promises to customers.
• Rule of thumb: use AI where a person can check the result faster than doing the work from scratch.

Keeping a human in the loop

The safest way to use AI in a business is to treat everything it produces as a draft. A person reviews it, edits it if needed, and decides whether it goes anywhere. This is often called keeping a human in the loop, and it is the single most important habit for using these tools well.

Review matters because AI tools can be confidently wrong. They can misread a number, mix up two customers with similar names, or fill a gap with something that sounds plausible but is not true. These errors are usually easy to catch if someone looks, and expensive if nobody does.

Draw a clear line between suggesting and doing. It is fine for a tool to suggest a new customer record, a work order, or a paragraph of proposal text. It should not create records, send messages to customers, or change anything that matters without a person approving it first. If a tool can act on its own, ask how to turn that off.

Make the reviewer responsible, not the tool. When an AI draft goes out under your name, it is your work. Set the expectation with your team that "the AI wrote it" is not an explanation for a mistake.

Review does not have to be slow. For most drafts, a quick read for accuracy, names, numbers, and tone is enough. Spend more time on anything a customer will see or rely on.

In Servora, the assistant can propose creating customers, work orders, or assets, and a person must approve each proposal before anything is created.

• Names and contact details match the right customer
• Numbers, dates, and model or serial details are correct
• Nothing is promised that you have not agreed to
• Tone fits how you talk to customers
• A person approved it before it was created or sent

Good inputs

AI tools are only as good as what you give them. A blurry photo or a vague instruction produces a vague result. A clear photo and a specific instruction usually produce something useful on the first try. The habits here are simple, and they also make your records better for people.

For nameplates and labels, get close and fill the frame. Wipe off dust or grime first if you can. Avoid glare by changing your angle rather than using the flash. Make sure every character is sharp, especially the model and serial numbers. If the plate is damaged or partly hidden, take more than one photo from different angles. A PDF of a spec sheet or a clear scan of a document also works well.

For written instructions, be specific about what you want, who it is for, and how long it should be. "Write something about the repair" leaves the tool guessing. "Write two short paragraphs for a homeowner explaining that we replaced the condenser fan motor, why it failed, and what to watch for" gives it what it needs.

Include the facts you want used, and do not expect the tool to know things it was not given. If the proposal should mention a warranty term or a lead time, say so in the instruction.

When a result is off, adjust the input instead of starting over. Add the missing detail, correct the wrong assumption, or ask for a shorter version. A second, more specific try is usually faster than editing a poor draft by hand.

In Servora, uploading a photo or PDF of a nameplate extracts the serial and model identifiers, and AI can generate or rewrite proposal text from an instruction you write.

1. 1Take one nameplate photo the way you normally would, then retake it close, sharp, and glare-free. Compare.
2. 2Write a one-line instruction for a proposal paragraph, then rewrite it to include audience, length, and the key facts.
3. 3Use the more specific version for a week and note the difference.

Data and privacy questions to ask any vendor

When you use an AI feature, your data goes somewhere to be processed. Customer names, addresses, equipment details, notes, and photos may all be involved. You are responsible for that information, so it is worth asking a few direct questions before turning any AI feature on.

First, ask whether your data is used to train models, either the vendor's own or a third party's. Many businesses are uncomfortable with customer information becoming part of a general-purpose model. A clear answer should be in the vendor's privacy policy, not only in a sales conversation.

Second, ask how your data is kept separate from other customers of the same vendor. Your records and anything the AI produces from them should stay within your account.

Third, ask whether AI features are optional and who can turn them on. You should be able to decide which features your team uses, and to leave them off if they are not a fit.

Fourth, ask what the AI can do on its own. Can it create records, send messages, or change data without approval? The answer should match how much control you want to keep.

Finally, ask what happens to your data if you leave. You should be able to get your records out in a usable form.

Get answers in writing, read the privacy policy yourself, and revisit the questions when the vendor adds new AI features.

In Servora, optional AI features are scoped to your tenant, and the privacy policy states that tenant data is not used to train third-party models.

• Is our data used to train any models?
• How is our data kept separate from other customers?
• Are AI features optional, and who can enable them?
• Can the AI create, send, or change anything without approval?
• How do we export our data if we leave?
• Where is this written down?