Mentat
Menu

The mechanism

From a voice memo to a context-rich estimate in the system you already run.

A technician records a short voice memo about what they saw on the visit. Mentat transcribes it, structures it into priced recommendations that carry the field context, and drafts the estimate back into the system the company already runs, where a person reviews each one and sends it. That is the whole loop, and the mechanism behind raising average ticket without the hard sell.

The loop

Four moves, one decision.

  1. Capture

    The tech talks for a few seconds, the way they already would to a dispatcher. No app to learn, no form to fill.

    Mentat technician mobile app showing a scheduled job with briefing cards of prior recommendations for the property and a tap-to-record button.
    The tech picks the job, taps record, and talks. What they need to know walking in is already on the screen.

  2. Structure

    Mentat transcribes the memo into priced recommendations, each a specific work item with the context behind it, what the customer said, and how time-sensitive it is.

  3. Review

    A person sees each recommendation with its full context and decides: quote it now, hold it for the season, route it to follow-up, or pass. Nothing moves toward a customer until they do.

    Mentat operator review queue showing a priced furnace replacement recommendation with priority score, provenance details, and the source voice memo transcript.
    The memo comes back structured: priced recommendations, priority, field context, and the technician's original words, waiting on one human decision.

  4. Draft

    The recommendation is written back into the field service system as an estimate with that context attached, so whoever picks it up sees the story, not just a number.

    A draft quote in Jobber created by Mentat, with a line item carrying the technician's field observation and a note attributing the source memo.
    The approved estimate lands in Jobber as a draft, field context and attribution carried through. A person still sends it.

Is it safe to put AI on customer-facing estimates?

Yes, because the AI never faces the customer. Mentat drafts, a person reviews, a person sends. The model proposes; it does not decide and it does not talk to anyone. And your data stays yours. Mentat keeps each company's data isolated and does not train on your customers' information, so no company's data ever sharpens a competitor's tool. The work stays inside your four walls.

Does it work with the system we already run?

Yes. Mentat sits on top of the field service system you already use and drafts the estimate back into it. It replaces nothing, and the tech learns no new app. There is more on which systems are live today and which are on the way.

What happens to everything the techs notice, over time?

It builds into a record of the property. Every captured visit adds to what Mentat knows about that site, so the next technician arrives with the history instead of starting cold. The coil someone flagged last spring. The upgrade the customer said they would think about. The part that was on its last season. On day one the value is the loop catching work that used to slip away. Over time it is a briefing that walks back in with the next tech.

It closes attributable, not on faith.

The loop earns revenue on the visit you are running now, and the record it builds makes the next one sharper. One thing finishes the loop: Mentat tracks whether the drafted work sold, so the revenue it recovers traces back to the recommendation that surfaced it and the tech who spotted it. It is attributable, not a number taken on faith.