I built a family scam alarm with AI agents on Switch. Here's how, and everywhere I got stuck
Most "AI agent" demos have one agent answering one person. The problem I wanted to solve needed the opposite: several people and several agents, in one chat, getting one urgent job done together. This post walks through how I built it, for the WeMakeDevs × Switch "Flip the Switch" hackathon . For each part I explain why it's built that way, what I tried first , and where it broke . If you're…
The report focuses on a family scam alarm system built using AI agents on the Switch platform. The system aims to detect digital arrest scams targeting older people and alert their families when such a situation arises. The design of the system avoids using AI for all tasks, instead delegating certain responsibilities to simple Python scripts. This approach ensures predictability, testability, privacy, and cost-effectiveness.
Key lessons learned from the project include:
1. Identify what should not be AI-driven: The system determines that certain tasks, such as assessing the urgency of a potential scam, should not rely on AI. Instead, rules-based approaches are used to ensure predictability and testability.
2. Treat non-AI services as participants: Switch treats Bank watch, a simple Python script, as another agent in the chat room. It has a name, identity, and room membership, allowing it to coordinate people, track tasks, and write complaints just like other AI agents.
The system can be built using Switch's open-source platform, which allows people and AI agents to interact in the same chat room. By using Switch as the foundation, the family can continue using their existing Telegram group while benefiting from the added functionality provided by the AI agents.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.