Mauro Genna

Four different user types — venues, advertisers, admins and internal messaging — needed AI assistance over the same production database, with radically different permissions and context. A single assistant with direct write access was not acceptable: a hallucinating model could corrupt billing records or leak data across accounts.

A role-based multi-agent system built on OpenAI function calling, where each role exposes its own scoped toolset. The core pattern is propose/execute: the model never writes to the database — it produces a structured proposal that a human confirms before execution. For admin reporting, a scoped read-only SQL sandbox allows arbitrary queries with zero write surface.

AI assistants running in production against real customer data with no unauthorized writes. The propose/execute pattern became the team's default for any AI-driven feature touching persistent state.

Billing and subscription data for Highpoint's customers lived entirely inside a third-party billing provider's own application — Captiview (Highpoint's ad-tech platform) had no direct ownership or visibility into that data. As the platform scaled, this became a growing dependency: critical customer records existed only in a third-party interface, disconnected from the product Captiview was actually building.

Architected an integration with the external payment processor's API to individualize and centralize billing records directly inside Captiview, rather than relying on that provider's application as the source of truth. The migration moved billing ownership in-house: Captiview became the system of record for every subscription, syncing full billing history for all existing customers into its own database instead of depending on a third-party interface.

Hundreds of active subscriptions integrated with zero disruption to existing customers. Captiview now owns and controls its subscription data instead of depending solely on the third-party provider's interface.

Residents reported incidents primarily by phone, generating a high volume of calls that staff often couldn't keep up with — leading to missed or delayed responses and no consistent record of what had been reported or when.

Built a voice-automation pipeline on Retell AI, integrated with backend services to capture structured incident data from resident calls and automatically trigger downstream workflows — email, SMS, and admin processes.

Automated incident intake end-to-end — every call is now logged with what was reported and when, without requiring a human to answer or transcribe it manually. Deployed across 50+ residential condominiums in Spain.