AI Voice Daily Reports: Speak Your Report (English or Spanish)

How it actually works

The foreman opens the daily report in the Field PM app and taps the voice option instead of typing. They talk through the day the way they would explain it to a super on the phone: who was on site, what got done, any quantities, anything that held them up, anything safety-related worth noting. Free browser-based speech-to-text handles the transcription (no per-minute cost), and then an AI parsing step — a few cents of Sonnet usage per report — reads that transcript and maps it onto the structured daily report fields.

Crucially, this is not a black box that submits on its own. The parsed report — crew list, hours by cost code, quantities installed, safety notes, weather and delays — comes back to the foreman as a filled-in form they review and correct before hitting submit. If the AI mis-heard a cost code or mixed up a quantity, the foreman catches it in that review step, the same way they would proofread a typed report before sending it.

What it captures from a spoken report

• •Crew and hours by cost code: "Three electricians on rough-in, cost code 16-100, eight hours each" maps directly to labor entries tied to the job's cost codes.
• •Quantities installed: "We pulled 400 feet of 4-inch conduit today" becomes a quantity entry against the relevant cost code, feeding the productivity-factor calculation automatically.
• •Safety notes and incidents: near-misses, toolbox talk topics covered, or PPE issues mentioned in the dictation get captured as safety-log entries.
• •Weather and delays: "Rained out from 1 to 3" gets logged as a delay reason, which matters both for productivity analysis and for any schedule-impact documentation later.
• •Equipment and materials: equipment on site and material deliveries mentioned in passing get pulled into their own fields rather than buried in a paragraph of notes.

Why English and Spanish both matter

A large share of field crews in the US, particularly in electrical, concrete, and mechanical trades, work more comfortably in Spanish than English, and a foreman who is fluent in the trade but less comfortable typing a long English-language report is exactly the person most likely to skip the report or hand it off incomplete. Building Spanish-language voice dictation directly into the same form — not a separate translated app — means the foreman reports in whatever language is natural, and the structured data that comes out (cost codes, quantities, hours) is language-agnostic on the back end.

This sits alongside Field PM's broader <a href="/features/form-library">Spanish field-facing rollout</a> — bilingual templates and calculators built for crews who work the field in Spanish even when the office runs in English. The goal in both cases is the same: match the tool to how the crew actually communicates, rather than asking the field to translate itself into the office's language before the data can be captured.

What this fixes in practice

The measurable problem AI voice reporting solves is adoption, not accuracy for its own sake — a report that never gets filed is worthless no matter how well-designed the form is. Contractors typically see daily-report completion rates climb noticeably once foremen can dictate in under two minutes standing in a truck at the end of the day, versus a 10-15 minute typing session that gets postponed, forgotten, or done from memory the next morning (with the accuracy loss that implies).

It also changes what shows up in the report. A foreman typing tends to write the minimum — "normal day, no issues" — because typing more takes more effort. A foreman talking tends to actually narrate the day, which surfaces the small delays, near-misses, and quantity notes that a rushed typed report would have skipped entirely. More complete daily reports mean a more accurate <a href="/features/pm-dashboard">real-time job cost picture</a>, which is the entire point of collecting the data in the first place.

The honest limits

Voice-to-form works best for the fields it is designed to extract — crew, hours, cost codes, quantities, safety notes, delays. It is not a substitute for careful photo documentation, detailed sketches, or attaching a signed form that requires an actual signature — those still need their own step in the app. And the review step is not optional busywork: cost codes especially can be misheard or ambiguous ("one-six-one-hundred" vs. "one-six-two-hundred"), so the foreman confirming the parsed report before submission is what keeps the job cost data trustworthy.

The cost is genuinely low — free browser speech-to-text plus a few cents of AI parsing per report — which is why it is included rather than sold as a metered add-on. The bigger win is behavioral: reports get filed the same day, in the language the foreman actually thinks in, with more detail than a rushed typed version would have captured.