Skip to content
English
Contact Support
truleo.co
  • There are no suggestions because the search field is empty.

Best practices for 911 call analysis in TruAssist

Learn how to set up 911 call analysis so the quality-assurance (QA) scoring and briefings match your agency's standards.

Use for: Dispatch supervisors and QA staff who review call-taker performance with TRULEO.

Why your SOP matters

TRULEO can only judge a call against standards it has been given. If the AI has no copy of your SOP, it falls back on general ideas of what a "good" call looks like. That can lead to results that don't match how your center actually works, such as:

  • A call that your team would call weak is scored as handled well.
  • A call-taker who repeats the same question is described as "diligent."

Loading your SOP gives the analysis a ground truth. Start there before you change anything else.

Step 1: Upload your SOP

  1. Ask an organization admin to open Document Management. (Non-admin users can't upload.)
  2. Upload your dispatch/call-taking SOP and any related policies (protocols, call-handling standards, QA forms).
  3. Wait for processing to finish before you run any analysis.

Tip: Upload the full, current version. If your SOP changes, upload the new version and remove the old one so the AI doesn't mix them.

Step 2: Have TruAssist build the prompt from your SOP

Instead of writing the QA prompt from scratch, let TruAssist draft it from your own document.

  1. Open TruAssist and confirm your SOP is available as a source.
  2. Ask TruAssist to turn the SOP into a 911 call-analysis prompt. For example:

    Using our dispatch SOP, write a QA prompt for reviewing 911 calls. Include the required steps a call-taker must follow, what a strong call looks like, and what should be flagged for supervisor review.

  3. Read the draft. Edit anything that doesn't match how your center works.
  4. If you already have a template, ask TruAssist to update it against the SOP and compare the two versions side by side. Pick the one that fits your QA process better.

Step 3: Write specific sections

Vague instructions produce vague results. Tell the AI what to look for, how much to write, and what to do when nothing is found.

❌ Too vague ✅ More useful
"Evaluate call-taker performance." "Check whether the call-taker obtained location, callback number, and nature of the emergency in the order our SOP requires. List each required step as Met, Partially met, or Not met, with a timestamp for each."
"Flag problem calls." "Flag any call where the caller reports a mental-health crisis, suicidal thoughts, or a medical emergency and the call-taker does not ask follow-up questions. Quote the exchange and give the timestamp."
"Note good work." "List up to 3 calls that showed strong calm-caller technique, with a one-sentence reason tied to our SOP."

Good habits:

  • Use fixed labels or rating scales (for example Met / Partially met / Not met) so results are easy to compare day to day.
  • Set counts and limits (for example "top 3 concerns") so the briefing stays readable.
  • Say what to write when nothing is found, such as "None identified," so you can tell the AI checked.
  • Ask for citations and timestamps so you can jump to the audio and verify.
  • Use your agency's own terms for call types, priority levels, and units.

Step 4: Account for how your center works

Tell the AI about anything that changes how a call should be read. For example:

  • Transfers and shared lines. If your center stays on the line while a call is transferred to fire or EMS, say so. Otherwise the AI may read another agency's call-taker as one of yours.
  • Calls you don't want scored, such as admin lines or test calls.
  • Which roles to evaluate (call-taker, dispatcher, or both).

Step 5: Set the briefing window and schedule

If you run 911 analysis as a recurring briefing, check:

  • How often it runs and what time window it covers. A schedule that runs every 12 hours may pull a shorter window than you expect, such as 7 days.
  • Time zone. Check that the run time matches your local time.
  • Who has access. Users need Analyst Briefings access to search, download, and analyze calls.

Step 6: Review and tune

  1. Run the briefing on a day you already know well.
  2. Compare the results to what your supervisors would have said. Look for false positives (praise for weak calls) and misses.
  3. When something is wrong, tell TruAssist what it got wrong and which SOP step applies, then ask it to revise the prompt.
  4. Confirm that flagged calls and accolades belong to your own staff.
  5. Re-check after any SOP change.

AI output is a starting point, not a final work product. A supervisor should review every flagged call before acting on it.

Quick checklist

  • SOP uploaded in Document Management and fully processed
  • Prompt drafted with TruAssist from the SOP and reviewed by a supervisor
  • Each section has specific criteria, a rating scale, and a limit
  • Transfer/shared-line handling described
  • Briefing schedule, window, and time zone confirmed
  • Right users have Analyst Briefings access
  • Test run checked against a day you know well