# Why Banks Building Their Own Compliance Agents Hit the Same Wall

DevFeed: [Why Banks Building Their Own Compliance Agents Hit the Same Wall](<https://devfeed.tech/articles/why-banks-building-their-own-compliance-agents-hit-the-same-wall-132562.md>)

Original publisher: [Read original article](<https://www.lyzr.ai/blog/why-bank-compliance-agents-stall/>)

Author: Lyzr Team

Published: 2026-11-13T07:50:45Z

Content type: article

Language: en

Sources: [Lyzr.ai](<https://devfeed.tech/sources/lyzr-ai.md>)

Topics: [ai observability](<https://devfeed.tech/topics/ai-observability.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [least privilege](<https://devfeed.tech/topics/least-privilege.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [audit](<https://devfeed.tech/topics/audit.md>), [shadow AI](<https://devfeed.tech/topics/shadow-ai.md>), [Amazon Bedrock AgentCore](<https://devfeed.tech/topics/amazon-bedrock-agentcore.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-compliance](<https://devfeed.tech/tags/ai-compliance.md>), [banks](<https://devfeed.tech/tags/banks.md>), [building](<https://devfeed.tech/tags/building.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [production](<https://devfeed.tech/tags/production.md>), [prototype](<https://devfeed.tech/tags/prototype.md>)

## AI overview

Banks' compliance agents often stall between prototype and production because they lack decision records, governed system access, agent identities and permissions, runtime policy enforcement, and a human review process that can handle volume. The excerpt says these are infrastructure and governance gaps, not model-quality problems. It reports no supervisory text requiring generative AI to be deterministic; the practical need is to reconstruct inputs, sources, model version, tool calls, approvals, and policy version. A shared governed path may reduce connector duplication, but it cannot reconcile inconsistent source data or remove organizational silos. Opencontroller is described as providing agent IDs and owners, per-call authorization, limits, approvals, signed audit trails, and reversible quarantine; it does not build agents or resolve those data and organizational problems. The suggested path is to inventory and tier agents, define and enforce each agent's permissions, route traffic through a policy point, test decision replay, set approval and sampling rules, and involve governance teams at design time. The excerpt cautions that vendor capability claims should be confirmed.

## Source excerpt

Banks that build their own AI compliance agents usually stall between a working prototype and production because the prototype proves a model can do the...