A clipped call sign can be enough to confuse a machine. On a security radio channel, that small mistake can mean a missed acknowledgment, a status that never changes, or an officer having to say the same thing twice. Privateer has spent more than three months teaching its AI dispatcher to handle the way people actually talk on the job.
By September 25, the system had accumulated 29,120 transmission records across 105 calendar days. Officers have been using it for more than 90 days, and Cody Betzer reports few to no complaints from the field. His assessment of its current operation is that it works almost flawlessly.
That confidence comes from using it through ordinary working days. The record behind it shows thousands of successful radio responses, regular status handling, and a review process that turns misheard speech into something the system can recognize next time. It also shows where the machinery still trips.
AI Dispatch is a separate add-on that integrates with Full Sail. Full Sail supplies the operational context the dispatcher needs, including assignments and duty status. AI Dispatch connects that information to the radio conversation, allowing an officer’s spoken report to produce an acknowledgment, update a status, or bring a matter to a human dispatcher’s attention.
The radio becomes a working interface
An officer can call Dispatch over Zello, make a radio check, report an operational status, or answer an hourly check-in. The system evaluates the transmission, identifies the sender, and checks the relevant working context before carrying out a routine update. When a reply is appropriate, it sends a spoken response back over the channel.
That connection matters because a radio acknowledgment and an operational record serve different purposes. The officer needs to know the call was received. The people coordinating the shift need a current picture of who is working and what they are doing. AI Dispatch can handle both sides of a routine exchange.
Hourly check-ins give it another recurring job. It tracks responses, follows the progress of the check, and can announce completion. It also handles clock-in reminders when a scheduled operative is transmitting without the corresponding duty record. The workflow keeps those reminders from needlessly interrupting an active check-in.
The application also supports spoken vehicle-information requests and follow-up questions. An officer can ask Dispatch to check a plate and receive available vehicle details over the radio. Those requests depend on the information available to the connected services; they do not turn the dispatcher into a police records system.
Safety-critical traffic receives different treatment. The system can flag a transmission and create a prompt requiring human attention. A supervisor still has a consequential job to do: review the situation, make decisions, and coordinate the response.
The lesson starts with a second listen
Privateer built a radio traffic review workspace around a practical question: what did the officer actually say?
A reviewer can select a recorded transmission, play the audio, slow it down, and drag across its waveform to isolate a difficult section. The automatic transcript sits beside a field for the corrected wording. The reviewer can listen to the same fragment until the words are clear, then save the correction.
Those corrections feed back into later processing. Approved wording can replace a recurring misheard phrase, or correct a matching transmission before the dispatcher interprets it. The original transcript remains available alongside the corrected version, so the change can be examined afterward.
By the time of this review, Privateer had saved 73 reviews, with 47 corrections approved for future use. Later transmission records show approved corrections being applied again. The system is retaining lessons from the radio traffic it encounters.
A person supplies the correction. AI Dispatch carries it forward when matching traffic arrives. That gives Privateer a way to improve recognition of its working vocabulary without asking officers to live indefinitely with the same mistake. The learning is specific: a corrected phrase improves that recurring case, while an unfamiliar error still needs someone to catch it.
Development work has followed the same practical direction. Punctuation inserted by speech recognition and words that sound like numbers can change how a call is interpreted. Privateer has refined those cases while preserving the distinction between an officer reporting their own status and talking about somebody else. Getting the words approximately right is only the beginning; the system also has to understand whose business those words describe.
What the working record shows
The most recent 30-day period examined, ending September 25, contained 5,323 transmission records. Of those, 1,952 recorded a successful broadcast response and 20 recorded a failed reply. Another 101 recorded an hourly check-in completion being sent. Across the full set, 36 transmissions carried an error record, including the failed replies.
For the 1,952 transmissions with a successful broadcast response, the median time from the application receiving the transmission to finishing its processing was about 9.5 seconds. Ninety-five percent finished within about 13.8 seconds. That interval includes the work of sending the spoken response; it is not a measurement of when an officer first heard the voice.
Those results help explain the positive field assessment. They also leave room for the failures that still need attention. A recorded successful broadcast establishes that the sending system reported success. It cannot establish that every word was understood correctly or heard clearly by everyone on the channel.
Silence needs its own explanation. In the same period, 1,894 transmissions were deliberately held from routine action by the system’s checks. Common reasons included uncertain transcription, traffic that was not addressed to Dispatch, and a sender whose duty context could not be verified. Officers talking to one another should be able to finish their exchange without the AI joining in.
That restraint is part of making the dispatcher useful. A system that answers every sound on the channel would create more work for the people using it. Privateer’s review tools make it possible to examine a silent response and decide whether the system exercised appropriate restraint or missed something it should have handled.
The work between one transmission and the next
Radio dispatch exposes an automation system to a succession of small, consequential decisions. It has to recognize a call, associate it with the right person, preserve the current status, speak when needed, and leave an intelligible record. A fluent voice is only one piece of that work.
Privateer has learned to give failed and interrupted workflows their own treatment. The application includes recovery for unfinished hourly check-in closeouts and deferred reminders. Before a delayed reminder goes out, it checks whether the reminder still applies. That matters on a shift where a person’s status can change while a service is recovering.
The radio review process adds another kind of recovery. A misheard transmission can become a correction that survives the shift. Over time, the people supervising the system build a record of what it gets wrong and a practical way to address recurring errors.
More than 90 days of officer use have given Privateer something substantial to work with: an AI dispatcher operating alongside its staff, a body of radio records, and feedback from people who depend on the channel. The next improvement can begin with an actual transmission—replayed, understood, corrected, and carried into the next day’s work.