DIGITAL TRANSFORMATION If an agent executes a trade or flags a transaction, that action must be entirely auditable, explainable and reversible.
In financial services, autonomy without accountability doesn’ t work so observability is foundational, not optional,” Bill stresses.
“ Leading firms are building full audit trails for every agent action – what data was used, what reasoning occurred and what decisions were executed.
“ As AI becomes more embedded in operations, governance shifts from a set of controls to a continuous system – spanning security, compliance and agent behaviour in real time.”
To achieve this level of transparency, control planes must be baked into the enterprise layer rather than treated as an afterthought.
“ That’ s why governance layers are embedded directly into the agentic stack, enabling explainability and reversibility from day one,” Bill says.“ At Microsoft, we think about this through the lens of Agent 365, bringing enterprise-grade identity, security, compliance and monitoring into how agents are deployed and managed, so every action is policy-aware and fully traceable.
“ As a control plane, it delivers endto-end observability for cloud and local agents, and it can secure agents and interactions. It can also discover and mitigate the risk of‘ Shadow AI’ agents.”
The three phases of AI maturity For a CTO at a tier-1 bank, navigating this landscape requires a roadmap. Using the classic crawl-walk-run framework,
technologymagazine. com 119