It's a Tuesday in March and a number you half recognize lights up your phone. The woman on the other end says the dishwasher you hooked up last spring is leaking at the connection, and wasn't there some kind of warranty on that? You tell her you'll check and call her right back.
Then you go looking. The invoice is in your email somewhere, or maybe the paper copy is still in the truck. You can't remember if you quoted a one year workmanship guarantee or if that was the job over in Needham. The part might be under the manufacturer's coverage, but you'd have to dig up the model number to even know.
That gap, between the promise you made and the paperwork that proves it, is where trades quietly lose money and goodwill. Honor a callback you didn't owe and you eat the labor. Miss one you did owe and the customer tells three neighbors. AI won't make the callbacks for you, but it's genuinely good at the part you hate: reading back through your own records and pulling every date and promise into one place you can actually see.
Every promise outlives the invoice
Think about everything you commit to on a normal job. Most of it never makes it onto a tidy list. It lives in your head, scrawled on the back of a work order, or buried in a text thread.
Here's the usual lineup:
- Workmanship warranties. The year (or two, or five) you stand behind your own labor. The clock starts the day you finish, and you're the only one tracking it.
- Touch-up promises. "I'll come back and caulk that corner once it cures." Small, easy to forget, and often the thing a customer remembers most.
- Manufacturer coverage. The water heater, the compressor, the shingles. Different parts, different terms, and the homeowner always assumes you'll know them.
- Maintenance you recommended. "Have us flush this in the fall." A promise and a future job, both at once.
- Seasonal callbacks. The deck you sealed in June that'll want another look before next winter.
None of these show up on the invoice as a date you can sort by. That's the whole problem. The promise and the proof live in different places, and neither one pokes you when it's time.
Get it into one list without retyping everything
You don't need new software with a monthly fee and a login your crew forgets. You need your own records read back to you and organized. That's a job today's AI does well, because it's reading and sorting, not judgment.
Start by gathering what you already have for a batch of recent jobs. Invoices, quotes, the notes app on your phone, the voice memo you left yourself driving home. Then hand them to an AI tool and ask it plainly:
- "Read these invoices and list every warranty or service promise you can find, with the customer, the job, and the date it was made."
- "From this voice memo, pull out anything I promised to come back and do, and when."
- "Put it all in a simple table I can paste into a spreadsheet, sorted by the date each promise expires or comes due."
A minute of dictation beats a form every time. If you talk to your phone at the end of a job ("finished the Marino bathroom, one year on the labor, told her I'd re-caulk the tub in two weeks, the fan's under a two year manufacturer warranty"), the AI can turn that run-on into three clean dated entries.
What a good promise tracker actually holds
Keep it boring and keep it short. A promise you can't find fast is a promise you won't honor. Five or six columns cover almost every trade:
- Customer and address. So you're not guessing which Sullivan it was.
- What you promised. One plain line. "One year workmanship," not a paragraph.
- Date made and date due. Both of them. The due date is what you'll sort and remind on.
- Who's on the hook. You, or the manufacturer. This is what decides who pays.
- Proof. A link or a quick note of where the invoice or photo lives.
- Status. Open, scheduled, or done.
The last column matters more than it looks. When a customer calls, you want to open one list, see the promise, confirm it's real, and know in ten seconds whether it's yours to fix. The awkward scramble turns into a quick scheduled visit.
Reminders that reach you before the customer does
Lists nobody opens are just shoeboxes with better handwriting. The point is to get poked at the right moment, not to go hunting for it.
Once your promises have due dates, you can set reminders a step ahead of each deadline. The AI can draft them; your calendar or phone delivers them.
- Warranty coming up on a year? Ask for a reminder 30 days out, so you can offer a quick check before it lapses instead of after something fails.
- Touch-up promised in two weeks? A nudge at ten days gives you room to fit it into a route you're already driving.
- Seasonal work? One reminder in early fall for everything you said you'd handle before the first hard freeze.
Batch the small ones. If three customers across town all need a quick return visit this month, one afternoon handles all three. A promise tracked becomes a promise scheduled, and often a second job you'd have otherwise let slide.
Where this falls short, and what to double-check
Be honest with yourself about what the tool can and can't do. AI reads and organizes. It doesn't know your trade's fine print, and it will sometimes get a date or a term wrong.
Watch for a few things:
- Warranty terms. If an invoice just says "limited warranty" with no length, the AI might guess. Don't let it. Tell it to flag anything unclear instead of inventing a date.
- Manufacturer coverage. The AI can note that a part is covered, but the real terms live in the manufacturer's own paperwork. Treat its entry as a pointer, not the final word.
- Your actual words. A promise is a real commitment, so read the list over once before you trust it, the same way you'd check a quote before sending it.
- Private details. Customer names and addresses are sensitive. Use a tool your shop trusts and keep the master list somewhere you control.
Not one of those is a reason to skip it. A list that's 95 percent right and reviewed by you beats a perfect system that lives in your head and fails the week you're slammed.
Worth doing this week
No need to boil the ocean. Start with the promises that are still live right now.
- Pull the last three months of invoices and jobs into one spot: an email folder, a drive folder, or your photos.
- Ask an AI tool to list every warranty and service promise it finds, with customer, date made, and date due.
- Drop the result into one spreadsheet and fix anything that looks off. Ten minutes, tops.
- Set one reminder for the next promise coming due, so the system proves itself on a real callback.
- From here on, leave a 20 second voice memo at the end of each job and let the list grow itself.
Get this running and the forgotten callback stops being a gut punch and turns into a line on a list you glance at Sunday night. If you'd rather have someone set the whole thing up around how your shop already works, that's the kind of thing we do at Massachusetts AI Agency. Either way the goal's the same: every promise you made, in one place, reminding you before the phone does.