Notes from our AI practice
Articles on putting AI to work in regulated industries, written by Swapnil Tamse, AI Practice Lead, Infinity AI.
-
The OCC's AI framework and the problem of agents that spawn agents
OCC Bulletin 2026-54 tells examiners to rank bank AI by consequence, autonomy and human oversight. It leaves agent swarms and hard limits open. Here is what banks can do now.
-
Improve the skill, not the model: an agent design for regulated work
Put the expert's method in a skill the business owns, put every number in a tested tool, and let the agent improve by proposing edits to the skill that a person approves.
-
From policy to control: measuring AI behavior in regulated finance
An AI policy says the model will be fair and grounded. A model risk officer asks for the number and the evidence. This is how a behavioral principle becomes a KPI with a threshold and an audit trail.
-
Turning many voices into one AI roadmap
A three-step roadmap for any organization building an AI program: executive education, a CTO roundtable and white paper, and an AI readiness scorecard that ends AI theater.
-
Why AI model validation needs fair lending bias testing
SR 11-7 was written for stable, explainable models. AI models drift and find proxies nobody programmed, so validation without bias testing leaves a fair lending exposure.
-
Three questions banks ask before deploying AI
Financial services requires three answers before any AI system ships. They come from banking regulation, and they apply to any software that makes decisions on someone's behalf.
-
The AI governance role your framework assumes you already have
NIST names an independent AI reviewer role, the TEVV specialist, that most organizations have never heard of. Adopting the framework is easy. Building the people it assumes is the hard part.
No articles match. Try another word or show all articles.