The note (verbatim).

Here's a real-shaped voice note from a rep pulling out of a customer's lot. Total recording length: 34 seconds. The rep talked the way they'd talk to dispatch, hands-free, while driving to their next stop.

That's the entire input. No typing, no fields, no dropdowns. The rep didn't pre-think structure — they just talked. By the time they reached their next stop, Voze had pulled the note apart and written the pieces to the right places.

What Voze extracts.

From those four sentences, Voze identifies and structures eight distinct pieces of data. Each one gets validated against your DMS account list, contact list, and parts catalog so the structure has real handles.

• Account: "Acme Trucking" — matched against the DMS account record (account ID, region, owner rep).
• Contact: "Sam" — matched against contacts known at Acme Trucking. If multiple Sams exist at the account, Voze flags for disambiguation rather than guessing.
• Competitive signal: "Samsara" — tagged as a competitor mention. Frequency tracked: third Acme mention in 60 days, which Voze flags as a pattern, not a one-off.
• Product signals: "brake suppliers" (category-level interest), "tires on units 4 and 7" (specific SKU need against fleet records).
• Action signals: Pricing discussion requested before quarter close — tagged as both an action item and a deal-timing signal.
• Follow-up: Thursday meeting — calendared on the account with the right person assigned.
• Deliverables: "Send the lane card and the multi-unit pricing" — captured as two specific follow-up tasks tied to the visit.
• Visit metadata: Timestamp (10:47 a.m.), location (Acme's address inferred from GPS), rep ID. Voice note kept alongside transcript for context.

Where each piece lands.

Extraction is half the story. The other half is where the structured pieces go — because data that doesn't end up where someone uses it is data that doesn't matter.

On the rep's account view (Voze Pro mobile).

Acme Trucking moves to the top of the rep's "this week" list. The note shows under the account record. The two deliverables (lane card, multi-unit pricing) appear as tasks with a Thursday deadline.

On the manager's territory feed (Voze Web).

The Samsara mention shows up in the competitor velocity view — third Acme mention in 60 days bumps Samsara's territory ranking. The action-flag layer fires an Urgent flag because the pricing conversation has a hard deadline tied to quarter close.

In your DMS.

The transcript writes to Acme's customer record as a note or activity (depending on platform). The "tires on units 4 and 7" tasking writes to Acme's parts/service record so dispatch and inside sales see it. The Samsara competitor tag writes to a custom field on the account record so it's visible to anyone opening Acme even outside Voze.

In your reporting layer.

The note rolls into territory aggregates: Samsara mention rate by week, brake-supplier interest by account, fleet-tire urgency by month. These become the patterns that drive Friday-afternoon decisions — see the field-visibility guide for how managers actually use them.

What it won't do.

Worth being clear about what Voze doesn't claim from a voice note, so nobody's surprised in week two.

It won't make up data the rep didn't say.

If the rep didn't mention a price, Voze won't invent one. If they didn't say a contact name, Voze flags for disambiguation rather than guessing. This is on purpose — false structure is worse than missing structure, because it erodes trust in the whole capture layer.

It won't close opportunities in your CRM automatically.

The voice note creates the signal. Whether that becomes a quote, a pipeline change, or a stage progression is a rep or manager decision. Voze surfaces the trigger; the human does the action.

It won't replace a sales process you don't have.

If your team doesn't have a follow-up rhythm today, Voze captures more signals than they'll act on. The capture step is solved; the action step is still on the team. That's why pilot success depends on the believer manager actually running coaching off the captured data, not just letting it accumulate.