AI architecture authoring and review

The challenge

Every new solution at the bank needs an architecture: written, drawn, checked against the bank’s standards, then reviewed for cloud, data, AI, security, resilience and cost. The knowledge needed to do that well was spread across many systems and many people, and reviews were slow and uneven.

My role

• Vision and solution design: defined the target, the use cases and the roadmap for agentic AI in enterprise architecture, from writing an architecture to reviewing it, with the bank’s enterprise architecture, security and AI teams.
• Platform architecture: governed integration with the bank’s knowledge repositories, alignment with its golden standards and technology positioning, deployment on its own cloud infrastructure and pipelines.
• Security and AI governance: controlled access to knowledge sources, experts kept in the loop, every recommendation traceable, explainable and auditable.
• Cost and quality: on my own initiative, a follow-up on small open-weights System One models to cut cost and improve quality.

The outcome

• Architects get a first, complete draft that already follows the bank’s standards, with non-conformities and risks flagged early.
• Reviews are consistent and auditable, and the cost of a review dropped by about 98%.