Prepared demos AI-assisted 911 platform with live transcription, locationing, translation and QA

Prepared, a vendor for public-safety communications, demonstrated an AI-assisted 911 platform during a July lunch-and-learn that presenters said is designed to give telecommunicators more real-time information and to reduce administrative burden.

At the session, Brad Flanagan and Bridal Flanagan of Prepared showed an “assistive call-taking” dashboard that integrates real-time transcription and summarization, spoken-location pin drops, device-derived location data (Apple/Android), streamed media and a library of call artifacts so supervisors and responders can access audio, video and transcripts from a single interface.

The presenters said the system ties into existing call-handling systems via a SPAN port so the transcription and summaries appear in real time rather than post-call. “We believe in, here at Prepared that we are moving into a new era of 9 1 1,” Brad Flanagan said, describing the product as an era of “assistive technologies” that return bandwidth to telecommunicators.

Why it matters: Prepared’s speakers said the platform aims to give call takers and supervisors faster situational awareness — for example, by plotting a caller’s spoken description ("7‑11 and First and Main") as a pin that evolves during a call, and by auto-generating a concise summary that can be copied into CAD notes. Presenters argued those features can speed dispatch handoffs, support evidence collection (media and audio are retained with transcript), and reduce duplicated work when calls are transferred between centers.

Key features demonstrated

- Real-time transcription and summarization: The interface shows caller audio, a running transcript and an automated English summary of the incident that supervisors can copy into CAD. Presenters noted the transcription can be corrected and that the model learns from feedback.

- Locationing: The product combines device-provided location (Apple/Android) and “verified spoken locationing” derived from the caller’s words; presenters said the spoken location pin updates as the conversation evolves.

- Media and retention: Call audio, streamed video and images can be captured and stored alongside transcripts. Presenters said the platform is browser-based and can align retention to agency policies so media and call artifacts are available without separate loggers.

- Keyword notifications and intents: Supervisors can set keyword alerts (for example: "fire," "shooting," "officer involved") and receive pings on phone or laptop to follow a call in real time. Presenters said notifications can be tuned (caller-only vs. caller-plus-call-taker) to avoid overload.

- Translation: Prepared Translator — previously Spanish-only — was shown expanded to support more than 21 languages, which presenters said covers roughly 96% of the non‑English–speaking population the company cited. The translation modes include SMS text-to-text, speech-to-speech/text and an AI bot that speaks caller responses when a call taker types the reply.

- Live protocol guidance and QA automation: Prepared demonstrated ingestion of an agency’s local protocols so the system can prompt call takers with the correct protocol questions in real time and automatically populate QA forms after calls. Presenters showed a supervisor dashboard with automated scores and a dispute workflow for call takers.

- Nonemergency triage with generative AI: Presenters described an AI-driven triage tool (not a traditional phone-tree) that they said has produced 40–60% reductions in nonemergency call volume in referenceable deployments. They emphasized the tool can escalate to a human when needed and can be configured to transfer callers into an agency’s workflow.

- Assistive dispatch / radio transcription: The platform can ingest radio traffic, transcribe it, summarize incidents and surface those summaries to call takers so they can verify radio reports without scrolling through long streams of audio.

Pilots, performance examples and caveats

Prepared said it is deployed in more than 1,000 PSAPs and works in all but one U.S. state, and that it serves large numbers of citizens through state contracts; presenters referenced a state-funded deal in New Mexico and ongoing partnerships there. They offered rapid-deployment examples: San Diego stood up the system in about 3.5 weeks, and Anoka in roughly 20 days to handle event-related calls. The presenters cautioned transcription accuracy is not perfect — they put it “north of 85, getting close to 90%” — and said agencies can flag errors (for example, pronunciation of “911”) so the model adapts.

During Q&A, an attendee identified as Jennifer asked whether the translated prompt “this is 911, what is your emergency?” sometimes came back to callers as “911” rather than the pronounced words. Prepared responded that the model learns over time and that staff can flag and correct specific translation errors to improve accuracy.

Data-sharing and transfers

The presenters described a development that conveys transcripts, audio and summaries over-the-top when a call transfers between centers so the receiving telecommunicator gets the contextual data immediately rather than starting from scratch; they cited New Mexico pilots for that feature.

No decisions or policy changes

The session was a product demonstration and discussion of features, pilots and configuration options. The presenters did not propose or adopt any formal policies or take votes during the meeting.

Next steps mentioned

Prepared invited attendees to follow-up demos at APCO in Baltimore and a Lehi 911 lunch-and-learn scheduled for Aug. 7; presenters said they will announce additional protocol-provider partnerships soon. Brad Flanagan and Bridal Flanagan encouraged agencies to request deeper demonstrations and to engage on pilots.

Ending

Prepared framed the platform as a single, unified repository for call, radio and media data designed to reduce manual work for QA and to provide supervisors and field responders with faster, consolidated incident information. Presenters stressed agency configuration — ingestion of local SOPs and tuning alerts — as a required step for real-world use.