Portfolio  /  Case study
Case Study · AI Process Automation

Gmail Money-Leak Agent

An AI agent that finds recurring charges, fees and dormant subscriptions without searching the whole inbox

AI Process Automation Personal finance
Gmail Money-Leak Agent project by Dilijent Systems
The Client

Gmail Money-Leak Agent is a Dilijent Systems applied-AI engineering project that finds hidden recurring charges, fees and dormant subscriptions in personal bank emails.

The challenge

Recurring charges, fees and dormant subscriptions hide in email statements, and finding them by hand does not scale. Giving an AI agent access to an entire inbox to run semantic search creates a bigger problem than it solves: those emails are among the most sensitive a person has, and an inbox-wide index is hard to scope, audit or delete.

Our solution

  • A Gemini plan-and-execute agent on Google Cloud Agent Builder, with a structured output fallback for bank emails in formats that are not yet known
  • No inbox-wide vector or hybrid search: per-bank templates plus Elasticsearch aggregations do the scanning
  • Recurring charge, fee and dormant anomaly scans strictly scoped by user_id
  • Gmail access through user_id-scoped OAuth, with multi-tenant Docker sandboxes for isolation
  • A human approves every finding before anything is acted on

The result

The agent finds recurring charges, fees and dormant subscriptions from the user's own mailbox, and every finding is explainable because it comes from an explicit scan rather than a model opinion. Per-user scoping means one person's mail can never surface in another person's results, and nothing is acted on until a human approves the finding.

— Contact Us —

Let’s talk about your business

Tell us what you need help with and our team will get back to you.

Send us a message

Fields marked * are required.

Your information is kept private.