Case Study - AI-powered crime reporting and security coordination

Overview

South Africa has one of the world's highest crime rates, with thousands of incidents left unreported. This is where Community Wolf comes in — an AI-powered WhatsApp chat interface that encourages reporting and keeps communities informed with local crime alerts. The founders saw the product's potential for the underserved security industry, where firms rely on WhatsApp to coordinate across hundreds of group chats. Seeing an opportunity to solve that, they approached us to partner on Enterprise Wolf: a suite of WhatsApp and AI-driven solutions tailored for security firms. The South African safety landscape is defined by fragmentation — public safety data is siloed, neighborhood-watch groups rely on disorganised communication channels, and private security firms operate in isolation. The information that would let any one of those groups respond faster is already flowing, but it's trapped in unstructured WhatsApp messages across hundreds of group chats, and no human dispatcher can read all of them in real time.

Working with the Community Wolf team, we deployed the first stage of Enterprise Wolf: group agents that listen into WhatsApp group conversations and extract crime, staff management reports, alarm activations, and other incidents in real time. The agents read each message, classify it against the incident schema, pull out structured data — type, location, time, severity — and trigger several upstream pipelines as a result. Verified incidents fan out to local communities through the same channel residents already use, so a reported break-in becomes an alert on a street one block over within minutes. Local security firms receive structured incident notifications that can route directly into their existing operations, replacing the human-reads-group-chat dispatch model. And staff management messages — shift handovers, patrol confirmations, alarm responses — feed into reports the firm can actually use, instead of being lost in the noise.

The AI pipeline converts unstructured community reports into actionable intelligence — which is the whole game in this domain. The agents have to be selective: a security WhatsApp group can carry hundreds of messages an hour, most of them not incidents, and a model that over-flags is worse than no model at all. The extraction layer is tuned for precision over recall in the live alerting path, with a separate review queue catching the harder cases for human triage. As the Enterprise Wolf suite expands, the goal is to transform how security firms operate — leveraging AI and WhatsApp to create a smarter, more responsive network that improves safety and efficiency across communities.

What we did