AI in Banking
Generative and agentic AI in financial services: delivery lifecycle, measurement, cost, model risk and the architecture that makes agents safe to run in a regulated bank.
The AI-Driven Development Lifecycle (AIDLC)
Six parts on using AI to build software in a bank: what it strains in the SDLC, how to measure it, how to verify AI-written code, what it really costs, how to control the tools, and the operating model that holds it together.
View the series →- 01AI does not break the SDLC. It breaks its pace and its gates
- 02Measuring AI-assisted delivery: from adoption to P&L
- 03Verification at AI speed
- 04The economics of AI-assisted delivery
- 05Controlling AI coding tools
- 06The AIDLC operating model: roles, gates and governance
Context Engineering for Agentic AI
Five parts on the discipline that decides whether AI agents are useful and safe in a bank: what goes into an agent's context, where it comes from, who is allowed to see it, and how to prove it worked.
View the series →- 01From prompt engineering to context engineering
- 02Retrieval and memory: architecting what an agent knows
- 03Tools, MCP and permissions: giving agents safe hands
- 04Context for regulated data: entitlements, PII and lineage
- 05Observing and evaluating agent context in production
All on AI in Banking
From prompt engineering to context engineering
An agent's quality is decided by what goes into its context window, not by the wording of its prompt. For a bank, that context is a governed data supply chain.
AI does not break the SDLC. It breaks its pace and its gates
Code written with AI is still deterministic. What strains is the lifecycle around it: human-paced sprints, review gates and requirements. The AIDLC redesigns them for a bank.
Before you switch on a vendor's AI agent: a five-step check
Core and technology providers are now shipping AI agents that act inside banks' systems. Here is what to establish before the first one goes live.
Fiserv's agentOS turns the core provider into an agent platform
The largest US core provider now wants to run banks' AI agents as well as their accounts. That changes the build-or-buy question for every community bank.
When the AI assistant becomes the front door to personal finance
OpenAI's purchase of Hiro is small. What it signals is not: the first conversation about a customer's money may soon happen outside the bank.
Turning Treasury's AI risk framework into a working programme
The new financial services AI risk framework has 230 control objectives. Here is how to adopt it without drowning in them.
AI agents will shop for your customers. Who vouches for them?
Card networks are building a trust layer for agent-led checkout. Banks that treat it as a network problem will lose the relationship it creates.