Voice to invoice: how to bill from the truck without typing.

The hardest part of invoicing on a phone is typing on a phone. Voice removes that friction. Talk for 20 seconds describing the work, the AI assembles the invoice from your price book and client records, you review the draft, and send. The client pays by card on a public link before you pull out of the driveway.

This guide walks the workflow end to end. Three setup steps, the actual voice phrasing that produces clean drafts, and when to override the AI vs trust it.

Voice-to-invoice is only as good as the price book it pulls from. Spend 30 minutes loading your common line items before recording your first voice memo. Three categories to cover:

• Service line items: trip charge, diagnostic fee, return visit, after-hours surcharge.
• Labor lines: hourly rate, half-day rate, journeyman vs apprentice if applicable.
• Common materials: 6-12 parts you reach for most. Wax rings for plumbers, fuses for electricians, filters for HVAC, paint by gallon for painters. The AI handles the long-tail materials from photo or text-add.

The first time you set this up, your price book is the foundation of every future voice draft. Time spent here saves time on every invoice for the next two years.

Open your invoicing app, find the voice button, hold and talk. The AI works best with plain trade language and a specific structure:

Examples of voice memos that produce clean drafts:

• Send Maggie an invoice for the Moen faucet job. About 2 hours labor, the faucet from the truck, and the wax ring.
• Send Carlos at 882 Beech an invoice for the panel inspection. Half-day labor and the diagnostic fee.
• Send Lakeshore Apartments an invoice for unit 4B. 4 hours labor, the disposal of the old unit, and the replacement condensate pump from this morning's pickup at Ferguson.

The AI fills in the rest from your price book and your client records. Quantities you didn't specify default to common values (1 hour minimum on service calls, etc.) which you adjust at review.

The AI shows you the draft invoice on screen. Three things to check, in order:

1. Client is right. AI matches names against your client list and can confuse 'Maggie at 14th street' with 'Maggie at Hawthorne Way' if you have two clients named Maggie. Pick the right one.
2. Line items match what you did. Quantities especially. If you said 'about 2 hours' and you actually worked 2.5, fix it before sending.
3. Total looks right. If the total seems off compared to similar past jobs, scroll back through line items to find what's missing or doubled.

Review takes about 10 seconds on a clean draft, 30 seconds on a complex one with multiple line items. Always faster than typing the invoice from scratch.

Voice and typing have different optimal phrasings. Three differences worth learning:

Use round numbers, not exact ones

'About 2 hours' beats '2 hours 17 minutes' because the AI rounds the latter awkwardly. You can fix exact times at review. Saying the round number lets the AI use the natural label from your price book.

Mention parts in plain trade language

'The Moen faucet' is enough if Moen faucets are in your price book. The AI matches the partial description to the line item. You don't have to say 'Moen Adler one-handle pull-down kitchen faucet, model 87233'. That works for typed invoices, not voice.

Skip the courtesies

Voice memos should be data-dense. 'Hi, hope you're doing well, just wanted to send you an invoice for...' adds noise the AI has to filter. Get straight to client name, job, line items.

The AI handles 90% of routine invoices well. The 10% that need override:

• Custom or bespoke line items not in your price book. Add them manually rather than asking the AI to guess.
• Pricing exceptions (discount you promised, friend-of-friend rate). Override the price book value and add a one-line note explaining the discount.
• Change orders mid-job. Always log these as separate line items labeled 'Change order' rather than expanding the original scope items.
• Time-and-materials jobs with significant variance. The AI rounds time; for hourly work where exact minutes matter, type the hours.

The full workflow, from job done to payment received, for a typical service call:

1. Finish the job. Walk back to the truck. Sit in the cab for 30 seconds.
2. Open the invoicing app. Hold the voice button. Talk for 20 seconds.
3. Review the draft on screen. Edit if needed (usually not). Tap send.
4. The client receives an email with a card-payable link. The link works on their phone, no login.
5. Drive to the next job. The client pays by card on their phone during your drive.
6. By the time you arrive at the next job, the payment has cleared. Stripe Connect routes funds to your account on the next business day.

Cycle time from job-complete to payment-received: typically 1-4 hours for clients with a card on file, same-day for clients paying for the first time, next-day for the rare bank-transfer client.

Voice-to-invoice processes audio of your speech. Two privacy questions worth answering before turning it on:

• Is the audio stored, and if so, for how long? Look for tools that transcribe and discard the audio rather than retaining it. If audio is stored, it should be encrypted at rest and deletable on request.
• Is the audio used to train AI models? If yes, walk away. Voice memos contain customer names, addresses, payment amounts, sometimes complaints about other contractors. None of that belongs in a training set.

Reputable invoicing tools (Falcon Bill, Jobber, Housecall Pro) all handle voice locally to the extent possible and use workspace-scoped models that do not train on tenant data. Always read the privacy policy before connecting any new tool.