ScanAI is a private system I built that listens to local public-safety radio, writes down what it hears, and keeps track of how well it is doing its job.
What ScanAI does
- Records each transmission as its own clip.
- Transcribes every clip into searchable text.
- Scores clip quality, so weak audio gets flagged instead of trusted.
- Tracks activity by town and by hour.
- Watches itself with a watchdog, so it keeps running unattended.
It currently holds more than 15,000 clips from six local towns.
Why I built it
I wanted to know what a small, careful AI system looks like when it has to run every day, not just in a demo. Radio traffic is about as messy as data gets: static, overlapping voices, clipped words and long stretches of nothing. If a system can make that searchable and measurable, it can handle your messy recordings, call logs or field notes too.
The ground rules
ScanAI follows the same rules as everything else I build. AI suggests, a person stays accountable, and every step leaves a record. A transcript is a convenience, never an official record, and the dashboard is shown with the transcript text blurred because the content is not mine to publish.
How to get access
ScanAI is a private tool. The live dashboard is at sai.jcody.com, and access is by request only. Contact me first, tell me who you are, and tell me what you would like to look at.
What it means for your project
The pattern behind ScanAI is simple: capture, transcribe, score, measure and monitor. It applies to call recordings, inspection audio, interviews and anything else that piles up faster than people can listen to it. If you are weighing AI development for your own organization, this is a working example with sensible limits. See the rest of my AI projects, read the field notes, or talk about your project.