MisaLabs built an AI agent that continuously checks whether a portfolio's composition is compliant with every UCITS rule - and approves new positions before they're traded.
UCITS compliance is a moving target: holdings shift through the day, the limits are intricate, and a single breach carries real regulatory consequence. The checks have to be exact, repeatable, and defensible after the fact.
The agent ingests portfolio files as they arrive, validates every holding against the UCITS limits with a deterministic rule engine, issues a compliance result at the daily cutoff, and routes new positions through pre-trade approval - all inside the firm's own tenant-isolated environment, with no portfolio data ever sent to external AI services.
How it works
The UCITS Compliance Agent runs as a pipeline of eight MisaCores - modular building blocks from our platform. Each one owns a single stage, from file intake to notification, and the whole pipeline runs inside the customer's own environment.
Foundational layer for secure storage, audit logging, event tracking, and AI model access.
Turns incoming portfolio files (CSV/XLSX) into clean, structured, validated rows.
Stamps each file with a SHA-256 checksum and provenance for full reproducibility.
Deterministic, versioned rule engine checking every holding against UCITS limits — unalterable by AI.
Processes files through the day, issues a result at the 5pm cutoff, handles late arrivals.
Review-and-sign-off state machine that re-triggers approval when the underlying data changes.
Renders the compliance digest and emails it to the recipient list.
Handles end-user authentication and session management for the platform.
