Design partner story · Banking
Giving regulators a provable answer to how AI agents are controlled.
A live design partnership focused on real developer workflows, real enterprise systems and measurable control.
IndustryBanking and financial services
RegionSoutheast Asia
Agents in useClaude Code and other tools
StatusDesign partnership active
The challenge
The risk arrived before the control.
The bank wanted its technology teams to benefit from AI agents. In a heavily regulated environment, though, ‘we trust the tool’ isn’t an acceptable control. The bank needed to show exactly what each agent could access, show that the controls couldn’t be bypassed, and produce evidence that stands up to audit and regulatory review.
What they’re doing with RayShield
- Running blast-radius assessments to show what each role’s agents could reach before anything runs
- Enforcing deterministic, default-deny policies: if it isn’t allowed, it’s blocked
- Keeping production credentials out of every agent’s hands
- Producing tamper-evident logs for the audit and risk teams
What success looks like
- A documented, testable control for AI agent use
- An evidence trail ready for internal audit and regulators
- Approved teams using AI agents on sensitive systems safely