What Is an AI System Deployer
A deployer under the EU AI Act uses an AI system under its authority. White-labeling, a substantial modification, or a change of intended purpose can make that organisation the provider.
Blog
Insights on AI for regulated industries.
Policy-driven agents, document intelligence, and enterprise automation — from the team building it.
Latest
Why AI Agent Savings Do Not Show Up in EBIT
McKinsey's 2025 State of AI finds 88 percent regular AI use and 62 percent at least experimenting with agents, yet only 39 percent report any enterprise EBIT impact, usually under 5 percent of EBIT.
Read article
All Posts
89 articles
A 12-Risk Go-Live Worksheet for an AI Agent
NIST AI 600-1 names twelve risks unique to or worsened by generative AI. File one owner-and-evidence block per risk, with residual-risk acceptance, before a tool-calling agent goes into production.
What Is Agentic Model Risk
Agentic model risk is the residual exposure after the revised guidance announced in OCC Bulletin 2026-13 placed generative and agentic AI out of scope, leaving agents outside the model definition.
Which Financial Services AI Controls Apply Before You Scale an Agent Pilot
Treasury released the FS AI RMF on 19 February 2026. CRI hosts 230 control objectives by adoption stage. Start with inventory, owners, data, oversight, vendor access, and an incident path.
The EU AI Act Clock Banks Are Using Is Already Wrong
Annex III high-risk duties under the EU AI Act now apply from 2 December 2027. 2 August 2026 still carries Article 50 transparency and enforcement, which trade pages keep getting wrong.
What Is Third-Party AI Risk
Third-party AI risk is the exposure a bank inherits when a vendor ships an agent, or when a core vendor turns on agent features inside an existing production release already under contract.
When the AI Agent Billable Unit Changes Mid Contract
Salesforce Flex Credits and Microsoft Copilot Credit packs can change the billable unit mid contract, so unused credits, org exclusivity, and a 125 percent disable cliff break the 2027 forecast.
Banks Lost Their Model-Risk Letter for Agentic AI. Examiners Will Still Ask.
OCC Bulletin 2026-13 and Fed SR 26-2 take generative and agentic AI out of model-risk guidance. Examiners will still ask how banks govern those tools while an AI request for information is pending.
AI Agents in Insurance Claims: Evidence-Linked Adjudication
How AI agents adjudicate insurance claims with evidence-linked decisions: document and photo intelligence, policy evaluation, exception routing, and audit trails that survive disputes.
The Real Cost of Drag-and-Drop Workflow Builders in 2026
Visual workflow builders demo fast and age expensively: canvas maintenance, branch explosion, diagram drift, and engineering ownership. The 2026 cost accounting, and the compiled alternative.
Why Your RAG Stack Cannot Pass an Audit
RAG retrieves context; audits demand provenance, determinism, and policy versioning. Why retrieval-augmented generation alone fails regulated decision workflows, and the architecture that passes.