Compl_ai

Inspiration

Existing tools like StarCompliance focus on helping compliance teams manage workflows, but nobody built for the employee experience. We wanted to flip that—build the ChatGPT of trading compliance. Something employees would actually want to use.

What it does

Compl_ai is an AI-powered compliance agent that gives financial employees instant, personalized trading guidance. Core functionality: Employee asks in natural language: "Can I trade Apple?" or "Can I buy Tesla this week?" System analyzes their profile: role (banker/trader/analyst), tier (1-4), active deals, research coverage, family conflicts Returns instant decision (<1 second): APPROVED, BLOCKED, or PRE-CLEARANCE REQUIRED Provides clear reasoning: "You're blocked because you cover AAPL in your research portfolio—FINRA Rule 2241"

How we built it

We quickly draw down the user interface and the backend flow. After the user puts in the request, the AI will extract that user's request and send it to the backend server. The back-end server will process the request using AI, which compares the user's request with the company restrictions and other details related to the user. Then we will tell the user if it's either approved, prohibited, or needs to contact a compliance officer.

We also implemented daytona in our app that when the user sends their request, we create a daytona sandbox to create real code of all the rules, so these companies can execute check deterministically and change policies automatically with the agent.

What we learned

1. Employee Experience Matters in B2B Financial software is notorious for terrible UX—complex dashboards, forms, jargon. We learned that bringing consumer product thinking (ChatGPT-style simplicity) to B2B compliance is a massive differentiator. Employees are users too.
2. Compliance is Complex Going in, we thought: "Just check if they're on the restricted list." 3 hours later, we understand why compliance teams exist—it's multi-dimensional reasoning with real consequences. This deepened our respect for the problem and our conviction that AI is the right solution.

What's next for Compl_ai

Phase 1: Complete the Core (Weeks 1-2) Audit Trail Implementation: Every query logged with timestamp, employee, decision, reasoning Full regulatory compliance (SEC 7-year retention) Immutable logs for audit defense Searchable dashboard for compliance teams Phase 2: Intelligence Layer (Weeks 3-4) Pattern Detection Agent: Detect suspicious behavior (repeated blocked queries) Identify MNPI leaks (coordinated queries about same stock) Flag high-frequency trading patterns Alert compliance proactively Self-Learning FAQ: Build knowledge base from common queries Auto-generate answers for frequent questions Reduce query volume by 60% Continuously improve

Built With

• flask
• python