How it works

From draft to defensible in four steps.

Report960 sits between your narrative and the file you turn in. It reads what you wrote, checks it against verbatim California code, shows you exactly where the report is soft, and lets you fix it before anyone else reads it.

Adds about two minutes to a report
01
Step one

Get the narrative in, however you work.

Report960 meets you where the report already lives. Dictate it in the car, upload the file your RMS spat out, snap your field notes, or type the draft straight into the editor with auto-save running.

Dictate

Speak the narrative and clean it up after.

Upload

DOCX and PDF, with bold and italics preserved.

Field note OCR

Photograph handwritten notes and pull the text.

Type a draft

Auto-saving editor, nothing to remember to save.

Report960 // intake
dui-narrative-2024-118.pdf
Uploaded · redactions preserved
PDFDOCXPASTEDICTATE
02
Step two

Analysis against verbatim California law.

When you run the analysis, Report960 retrieves the actual Penal, Vehicle, and Health and Safety sections that apply, then checks your narrative against the elements those sections require. Every finding has to point at real statute text.

Retrieval first

Findings are grounded in sections pulled from the codebook.

Quote anchored

If a quote is not word-for-word in your narrative, the finding is dropped.

Element checklist

See which elements of the charge your narrative supports.

Consistency check

Times, names, and vehicle details cross-checked across the report.

Report960 // grounded citation
CVC § 23612
Chemical test advisement
PC § 148(a)(1)
Element unmet
CVC § 21453(b)
Signal timing
Verbatim · CA Law Library
CVC § 23612(a)(1)(D)

“The person shall be told that his or her failure to submit to, or the failure to complete, the required chemical testing will result in a fine…”

Citation verified
03
Step three

Fix it inline, and let the score prove it.

Click any issue and the document jumps to the exact passage. Edit it in place. Report960 re-checks the revised text against the same statute and, when the defect is actually cured, marks the issue resolved and updates your readiness score.

AI re-verification

Edits are re-checked before anything is marked resolved.

Versioned revisions

Word-level history of every change you made and why.

Severity order

Critical defects surface first, informational notes last.

Keyboard flow

J and K to move between issues without touching the mouse.

Report960 // narrative
The driver seemed intoxicated. I observed red, watery eyes and an odor of alcohol
Revision savedv4 · tracked
04
Step four

Attest, then export a clean file.

Before anything leaves the system you attest that you reviewed the AI findings yourself. Export a PDF or Word file that matches what your agency already files, with the audit trail preserved on the record behind it.

Officer attestation

A required review block, not a checkbox buried in settings.

Soft lock

Editing after attesting clears it, so exports always match reviewed text.

PDF or DOCX

One click, formatted for filing.

Full audit trail

Who ran what, when, and what changed after.

Report960 // export
Narrative (final)
Citation list
Audit trail
Officer attestation
PDF DOCX
Boundaries

What it never does.

The limits matter as much as the features. These are the four things Report960 will not do, by design.

It never rewrites your narrative

Report960 flags, quotes, and suggests. The words that go on the report are yours, and every change is made by you in the editor.

It never cites law it cannot quote

Citations are matched against retrieved statute text. Anything the library cannot back up is dropped before you see it.

It never files or submits anything

Nothing is transmitted to your RMS, your agency, or a prosecutor. You export a file and take it from there.

It never replaces your judgment

Findings are a second set of eyes, not a signature. Every issue closes in one of three documented states - an AI-verified fix, not applicable under your agency's policy, or an officer override with your written basis. No silent dismissals, and the attestation records that the call was yours.

Please note: automated review is an assistive control, not an authority. It may overlook a defect or misjudge its significance, which is why each boundary above is enforced in the product itself rather than asserted in a disclaimer.

Capture

Get the report in however you work.

Dictation, field-note capture, upload, or paste - the review is the same either way.

Dictate the narrative

Talk the report out in the unit while it's fresh. Dictation drops straight into the narrative editor, ready to clean up.

Field notes by camera

Photograph a notepad page and the text is extracted into the draft - no retyping scene notes at the end of shift.

Or upload and edit

Drop a PDF or DOCX, or paste the narrative. Either way you edit in place, Google-docs style, with auto-save and revision history.

Quiet by default

Verification results, saves, and exports land in a notification center next to your credits instead of stacking popups over the report.

Policy-aware review

It only flags what your agency actually requires.

Body-worn camera rules aren't the same in every department, and they weren't the same five years ago. Report960 reviews each report against the policy that was in force - so officers stop dismissing noise.

Your requirements, not ours

Agencies set what they actually mandate - body-worn camera activation, supervisor notification, evidence logging. Checks your department never adopted simply don't fire.

Effective dates respected

Policies carry a start date. A 2019 report is reviewed against 2019 rules, so old cases and mutual-aid reports don't light up with requirements that didn't exist yet.

Documented, not deleted

A suppressed check is recorded as not applicable with the reason - it shows up in the export instead of quietly disappearing.

Auditable by design

Every finding ends in one of three documented states.

Nothing gets waved away. If an issue keeps failing verification, the officer can escalate it for a full-document re-check rather than guessing at what the model wants.

01
AI-verified fix

You rewrite the flagged passage and the AI re-checks that specific issue against the full narrative. It closes only when the language actually resolves it, and the projected score moves as you go.

02
Not applicable

The requirement doesn't apply under your department's policy or the date of the incident. Recorded with the reason, no judgment call needed from the officer.

03
Officer override

The officer disagrees - the AI misread the narrative, or the finding is wrong on the law. They pick a category, write a one-sentence basis, and it lands in the audit trail and the export as an officer judgment, not a silent dismissal.

Under the hood

Why the findings hold up.

Retrieval grounded

Analysis runs against statute text pulled from the California codebook, not model memory.

Quote anchored

Every flagged passage must appear verbatim in your narrative or the finding is discarded.

Versioned and auditable

Runs are timestamped and comparable, so you can show what the system saw years later.

Officer in the loop

Nothing is marked resolved or exportable without your review and attestation.

Ready when you are

See it on your own report.Sourced, quotable, traceable.

Run one narrative through Report960 and see exactly where it is soft before your sergeant or the DA does.