How it works
From a voice note to a draft in six steps.
You turn voice notes into estimates and invoices by having AI transcribe a tech’s recording, pull out the parts, labor and recommendations, price them from your own price book and build a draft in your estimating or accounting software for someone to check and send. The tech talks; the office checks; the typing disappears.
That matters because paperwork is where service businesses leak time and money. Estimates get typed at night from memory and scribbled notes, so they go out a day late. Small line items, like a filter or a shutoff valve, get forgotten. The office retypes what the tech already said out loud. A voice-note workflow fixes all three.
- 1
Record
The tech records a one-minute note in the truck, right after the job.
- 2
Transcribe
Speech-to-text turns the recording into written words.
- 3
Extract
AI pulls out the customer, the work done, parts, hours and recommendations.
- 4
Price
Each item is matched to your price book. Anything unmatched is flagged.
- 5
Draft
A draft estimate or invoice appears in your software, on your template.
- 6
Check and send
A person reviews the draft, fixes anything wrong and sends it.
An example
What a good voice note sounds like.
Here’s an illustrative note from an HVAC tech. It’s made up, but it’s the kind of note that drafts cleanly:
“Job at 412 Elm Street, customer Jensen. Furnace wasn’t igniting. Replaced the flame sensor and cleaned the burners, one hour of labor. Replaced the filter, sixteen by twenty-five by one. Recommend replacing the inducer motor, it’s noisy and close to failing. Quote that one separately.”
From that, the workflow builds an invoice with three lines: the flame sensor at your price-book price, one hour of labor at your rate, and the filter. It also builds a separate estimate for the inducer motor replacement, priced from your book, ready for the office to send with a note. Nobody typed a word.
Teach your techs a few habits and the drafts get cleaner:
- Start with the customer name and address
- Say numbers slowly, with units: “one hour,” “two fittings”
- Use the part names from your price book
- Say “quote separately” for anything recommended but not done
- Snap a photo of the nameplate for model and serial numbers
- Record in the cab, not next to a running compressor
Setting it up
Three ways to build it.
The draft needs to land where your office already works: your field-service platform or your accounting software. How you get it there depends on what that software already does.
Check what you already have
Some field-service and accounting tools have added dictation or AI drafting. Test it on ten real notes before you buy or build anything else.
A connected workflow
A transcription service, an AI step and your software’s connection, linked with an automation tool. Works well when your price book and templates are tidy.
A custom build
For multiple templates, complex pricing rules, several trades or software that needs a custom connection. Built and tested around your notes and your price book.
Get ready
What you need before you start.
- A current price book with part names and labor rates
- Your estimate and invoice templates
- Software that accepts outside connections
- Ten to twenty real voice notes to test with
- A named person who reviews every draft
- A rule for anything not in the price book
The price book is the one most shops underestimate. If half your parts are priced on the fly, the AI has nothing to match against, and every draft comes back full of flags. Cleaning up the price book first pays off twice: better drafts, and more consistent pricing across your techs.
Get it right
Keeping AI estimates accurate.
Transcription is good, but it isn’t perfect, and the mistakes cluster in predictable places. Numbers sound alike: “fifteen” and “fifty” are one bad connection apart. Units get dropped. Model numbers come out garbled. Trade terms and brand names get swapped for common words. Road noise, a running compressor or a customer talking in the background all make it worse.
Build the workflow to catch those errors, not hide them. Lines the AI isn’t sure about get flagged in the draft. Anything not in the price book gets a blank price and a flag, never a guessed number. Quantities that look odd, like fifty filters on a service call, get highlighted. And the draft always shows the original transcript next to it, so the reviewer can check what the tech said.
Most of all, a person checks and sends every estimate and invoice. An estimate is a promise to a customer, and an invoice is a bill; both need a human signature. The AI’s job is to get the draft close in seconds, so the check takes a minute instead of the whole evening.
Prove it first
Run a two-week pilot.
Pick one type of job, like service calls or maintenance visits, and one or two techs who’ll give honest feedback. For two weeks, they record notes and the office compares each draft with the estimate or invoice they’d have written by hand.
Track three numbers: how long from the end of the job to the estimate going out, how many corrections each draft needed, and how many office minutes each one took. Compare them with a normal week. If drafts need heavy rework, the fix is usually in the price book or the note habits, not the AI. Fix those, run another week, then expand to more job types.
What it costs
How Lithium builds it, and what it costs.
Lithium builds voice-note estimate and invoice workflows as custom AI automation . Each project is scoped after a free discovery call and quoted before any build starts; most custom builds take one to two weeks. Bring a handful of real voice notes, your price book and the name of your software, and a co-founder will tell you which of the three setups fits.
This is separate from our AI receptionist, which answers your phones for $179 a month. The two fit together: the receptionist books the job, and the voice-note workflow handles the paperwork after it.
Paperwork is one of several jobs worth automating in a service business. See five AI automations every home service business should adopt for the others, and the marketing glossary for any unfamiliar terms.
