LifeOps Financial Agent

Inspiration

Managing recurring digital subscriptions and utility bills is prone to silent price hikes and hidden fee increases. Most traditional bill-pay tools blindly execute auto-debits regardless of price changes, while fully autonomous AI agents risk taking financial actions without user oversight. We built LifeOps Financial Agent to give users an intelligent assistant that handles tedious bill auditing while maintaining absolute human control over financial transactions.

What it does

• Invoice Ingestion & Audit: Parses raw invoice text and receipts to extract vendor details, billing cycles, and current payment amounts.
• Price Anomaly Detection: Compares incoming bill amounts against historical baselines to immediately detect price increases and subscription tier changes.
• Human-in-the-Loop Guardrail: Automatically halts execution and triggers a safety gate when an anomaly is detected, requiring explicit user approval before any payment tool can be invoked.
• Safe Payment Execution: Interacts with financial tools to process payments only after human verification is confirmed.

How we built it

• AI Orchestration: AWS Strands Agents and AWS Bedrock AgentCore for intelligent tool routing and decision-making workflows.
• User Interface: Streamlit for a clean, interactive dashboard allowing real-time audit visualization and approval toggles.
• Core Backend: Python and Boto3 for custom tool integration (parse_invoice, check_price_anomaly, execute_payment).

Challenges we ran into

Building an agentic workflow that balances autonomy with strict financial safety was our primary challenge. We needed to ensure the agent wouldn't bypass human validation when detecting financial anomalies. We solved this by implementing a deterministic human-in-the-loop safety gate in Streamlit that intercepts tool calls before payment execution.

Accomplishments that we're proud of

• Successfully integrating AWS Strands Agents with custom Python financial tools.
• Designing a seamless Human-in-the-Loop (HITL) architecture that makes autonomous AI safe for real-world personal finance tasks.

What we learned

We gained deep experience orchestrating multi-step agent reasoning with AWS Bedrock AgentCore and structuring custom tools with AWS Strands Agents to enforce safe execution boundaries.

What's next for LifeOps Financial Agent

• Automated Dispute Generation: Automatically drafting cancellation or price-negotiation emails to vendors when price hikes are flagged.
• Multi-Currency & Bank API Support: Direct integration with open banking APIs (such as Plaid) for automated live bank sync.

Built With

• amazon-web-services
• bedrock
• boto3
• python
• streamlit