The State of Agentic AI Governance in US Financial Services 2026
Since revised US model-risk guidance placed generative and agentic AI outside its scope, no one has published data on what institutions are actually doing about it. This study measures the gap: which frameworks fill it, which controls are real, and what is blocking production — reported as the citable industry baseline.
By Evgeny Aleksandrov, Founder, BlackGrid · individual responses confidential
Who we're surveying
Risk, compliance, and model-risk leaders at US banks and credit unions
Data/AI and operations executives at lenders and payments firms
Wealth and asset-management technology and supervision leaders
What we ask (~7 minutes)
Where agentic AI actually sits today — exploring, pilots, or production, and in which workflows
How you govern it after the OCC 2026-13 / SR 26-02 carve-out — NIST AI RMF, ISO 42001, Treasury FS AI RMF, or nothing formal yet
Which controls exist in practice: reconstructable audit trails, human-in-the-loop thresholds, evaluation, reason codes, fair-lending testing
What is blocking you, and where 2027 budgets are headed
What participants get
The full report before public release
A benchmark cut: your answers against the aggregate
Anonymity by design — only aggregates are ever published
Free checklist
Join the research panel
Leave your details and we'll send the survey link when fielding opens for your segment — and you'll get the report before public release. Only aggregates are ever published.