As agentic AI systems move from pilots into real banking workflows, a new question is emerging inside financial institutions:
It’s no longer enough to say you follow AI governance principles. You must prove how your system behaves in practice.
At Unique AI, we see this shift firsthand. Our solutions are used by more than 30,000 financial professionals across 40+ institutions, including Pictet, Julius Baer, BNP Paribas, SIX Group, and Helvetia. As adoption of agentic AI accelerated, so did a shared concern from our customers’ risk, compliance, and AI teams:
How can we demonstrate with evidence that these AI systems are governed responsibly once they are live?
This is the gap we set out to close together with LatticeFlow AI.
A FINMA-Aligned Technical Blueprint
Together with LatticeFlow AI, we developed the first FINMA-aligned technical blueprint for assessing an agentic AI system already in production.
The assessment was conducted on our Investment Insights Agent, an AI system that helps relationship managers and client advisors generate personalized investment rationales and recommendations faster and more consistently.
Instead of relying on static documentation, the blueprint evaluates how the system behaves in real-world banking conditions across four dimensions:
- System reliability and robustness
- Explainability of outputs
- Human oversight and intervention mechanisms
- Ongoing risk monitoring over time
By mapping FINMA’s principles to concrete technical controls, including testing, monitoring, explainability analysis, and robustness checks, the blueprint turns governance from abstract guidance into decision-ready technical evidence.
What This Means for Financial Institutions
For banks and insurers, this blueprint provides:
- A concrete basis to evaluate and approve agentic AI systems
- Evidence to support continuous oversight after deployment
- A way to connect regulatory expectations with observable system behavior
- Documentation that is audit-ready, not interpretation-based
In short: governance grounded in real evidence.
Unique AI & LetticeFlow AI teams
Unique AI x LetticeFlow AI: Introducing FINMA-Aligned Technical Blueprint
As agentic AI systems move from pilots into real banking workflows, a new question is emerging inside financial institutions:
It’s no longer enough to say you follow AI governance principles. You must prove how your system behaves in practice.
At Unique AI, we see this shift firsthand. Our solutions are used by more than 30,000 financial professionals across 40+ institutions, including Pictet, Julius Baer, BNP Paribas, SIX Group, and Helvetia. As adoption of agentic AI accelerated, so did a shared concern from our customers’ risk, compliance, and AI teams:
This is the gap we set out to close together with LatticeFlow AI.
A FINMA-Aligned Technical Blueprint
Together with LatticeFlow AI, we developed the first FINMA-aligned technical blueprint for assessing an agentic AI system already in production.
The assessment was conducted on our Investment Insights Agent, an AI system that helps relationship managers and client advisors generate personalized investment rationales and recommendations faster and more consistently.
Instead of relying on static documentation, the blueprint evaluates how the system behaves in real-world banking conditions across four dimensions:
By mapping FINMA’s principles to concrete technical controls, including testing, monitoring, explainability analysis, and robustness checks, the blueprint turns governance from abstract guidance into decision-ready technical evidence.
What This Means for Financial Institutions
For banks and insurers, this blueprint provides:
In short: governance grounded in real evidence.
Unique AI & LetticeFlow AI teams