Fraud-prevention company Forter enabled about 200 technical and non-technical employees to build internal AI agents during a two-week program, according to a presentation by principal engineer Ben Maraney. The effort focused on making common capabilities reusable instead of asking every participant to solve retrieval, integration, and deployment from scratch.
Forter exposed company functions through custom Model Context Protocol servers, which give agents a standard way to discover and call tools. Participants could choose between no-code and code-based platforms based on their skills. The team also avoided complex retrieval-augmented generation setups where simpler access patterns worked, reducing the infrastructure participants had to understand before producing a useful workflow.
Security and legal teams were involved in defining acceptable access and use rather than reviewing every project only at the end. That organizational design is as important as the software: a short workshop can produce prototypes quickly, but production use still requires ownership, permissions, monitoring, and data controls. Forter’s experience is one company’s implementation, not a universal deployment recipe, yet it shows how shared tools and pre-agreed guardrails can shift an agent program from a specialist exercise to broader internal experimentation.