Categories
Explore all topics covered on AgenticAI Foundry
GenAI
Foundational concepts in Generative AI — from how LLMs work to RAG, prompting, and production deployment.
Evaluation & Observability for Production Agentic Systems: Metrics, Tracing, and Drift Detection Beyond the Demo
Production Agentic AI systems fail in ways that unit tests and demo environments cannot catch. A comprehensive framework for evaluation, tracing, and drift detection that keeps autonomous agents reliable at scale.
Token Economics and Cost Engineering for Enterprise GenAI at Scale
At enterprise scale, token costs compound fast. A technical deep dive into caching strategies, model routing, context compression, and the architecture decisions that can cut your GenAI infrastructure bill by 60–80%.
Small Language Models in the Enterprise: When SLMs Beat Frontier LLMs
Frontier LLMs get the press, but SLMs — fine-tuned, fast, and deployable on private infrastructure — are often the better choice for enterprise use cases. When and how to choose them.
Agentic AI
Deep dives into autonomous AI agents — orchestration patterns, multi-agent systems, and operational frameworks.
Model Risk Management Meets Agentic AI: Extending Three-Lines-of-Defence to Autonomous Agents
Traditional MRM frameworks were designed for static, batch models — not autonomous agents that evolve, chain actions, and interact with production systems. A framework for extending three-lines-of-defence to Agentic AI.
Defending Against Prompt Injection & Memory Poisoning in Multi-Agent Systems: A Banking Case Study
Prompt injection and memory poisoning are not theoretical threats in production multi-agent banking systems — they're live attack vectors. A technical defence architecture with a banking case study.
The Guardian Agent Pattern: A Safety Layer for High-Stakes Banking Actions
Before any autonomous agent executes a high-stakes banking action, a guardian agent should verify intent, check policy, and require human approval. A pattern that makes Agentic AI safe enough for production in financial services.
Agentic AI Platforms
Evaluation and architecture of enterprise Agentic AI platforms — AWS Bedrock, Palantir AIP, Azure Copilot Studio.
Agent-to-Agent Interoperability and the Emerging Agentic Commerce and Payments Stack
When AI agents transact with each other autonomously, payments infrastructure needs to evolve. A look at the emerging A2A commerce stack, identity requirements, and what this means for financial services architecture.
Building a Zero-Trust Agent Identity and Permissions Model for Financial Services
Every agent in a production multi-agent system needs a cryptographically-verifiable identity, scoped permissions, and a full audit trail. A technical blueprint for zero-trust agent identity in financial services.
Evaluating Agentic AI Frameworks for Regulated, High-Stakes Environments: LangGraph vs Microsoft Agent Framework vs Google ADK
A rigorous, criteria-driven framework evaluation for practitioners building Agentic AI in regulated environments — covering state management, auditability, security controls, and vendor lock-in risk.
Agentic AI Solutions
Real-world Agentic AI implementations in BFSI — KYC, fraud, core banking, payments, insurance, and more.
From Pilot to Production: An 18–36 Month Agentic AI Transformation Roadmap for Banks
Most Agentic AI pilots in banking never reach production — not because the technology fails, but because the transformation programme lacks the right structure. A detailed 18–36 month roadmap for banks that are serious about scaling.
Agentic AI and the EU AI Act: A Compliance Architecture for High-Risk Credit and Insurance Decisioning
The EU AI Act places Agentic AI systems used for credit and insurance decisioning in the highest-risk category. A complete compliance architecture — covering conformity assessment, data governance, audit trails, and human oversight.
Designing a Multi-Agent Architecture for Core Banking Modernization: Patterns, Pitfalls, and a Reference Blueprint
A practitioner's reference architecture for using Agentic AI to progressively modernize core banking — without the risk of a big-bang core replacement. Covers the integration layer, domain agent design, and the failure patterns that derail these programs.