Amazon and Salesforce Now Let AI Read Body Camera Video for Investigators
Hours of manual video review could shrink to one plain-English question.
Police departments, courts, and other government offices sit on mountains of unstructured evidence: body camera footage, surveillance video, photos, audio recordings, and scanned paperwork. Right now, humans have to watch, read, and tag all of it before anyone can act. Amazon and Salesforce just published a blueprint for automating that first pass, pairing Amazon's storage and AI data-extraction tools with Salesforce's Agentforce assistant (an AI that can take actions for you, not just chat).
Here's how it works in plain terms. Files land in Amazon's cloud storage. When a new file arrives, it automatically triggers an AI job that reads the file based on its type — transcribing audio and video, describing what's visible in images, pulling key fields out of scanned documents. The results get saved separately, ready to search. Then an investigator types a normal question into Salesforce, like 'show me footage from this address last Tuesday,' and the answer comes back inside the same window they already use. No switching apps, no watching 40 hours of tape first.
The pitch is speed. Manual review of body camera footage is one of the most time-consuming jobs in law enforcement, and it delays cases for victims and defendants alike. Automation could cut that from days to minutes, and it applies well beyond policing — think insurance claims, hospital records, or legal discovery. Salesforce and Amazon describe it as a flexible pattern, meaning other industries can copy the same setup.
The catch is significant. This isn't a product you buy and switch on — it's a technical recipe that requires cloud engineers, ongoing computing costs, and careful data governance. Automated transcription and image description can also be wrong, which matters enormously when the output could influence someone's freedom. Agencies will need human review, audit trails, and clear rules about who can query what.
- Amazon's AI reads video, audio, images, and scanned paperwork automatically, pulling out transcripts and key details so people don't have to.
- Salesforce Agentforce lets staff ask questions in plain English and get answers inside the same screen they already work in.
- It's a do-it-yourself blueprint, not a ready-made product — agencies need engineers, budget, and strict oversight rules.
- The same approach could work for insurance claims, hospitals, and legal document review.
Why It Matters
Could turn days of evidence review into minutes, but automated mistakes about real people carry serious consequences.