The 21 architectural patterns that underpin every production AI system — each mapped to PSF safety domains and PAI-8 governance controls, with enterprise examples, production failure modes, and implementation checklists. Written for practitioners who need to deploy, govern, and audit these systems.
The seven foundational patterns every production AI practitioner needs to understand. These appear in virtually every enterprise AI system.
Sequential task decomposition where each model output feeds the next input.
A classifier that directs each input to the most appropriate specialist agent or pipeline.
Running multiple agent tasks simultaneously and synthesising the results.
An agent critiques and revises its own output before it reaches a human.
The pattern that turns a language model from a text generator into an actor.
Specialised agents working together, each owning a domain of the overall task.
A controlling agent that directs sub-agents, manages state, and decides when a task is complete.
Seven patterns that determine whether a system survives contact with real users, real data, and real operational environments.
How agents store and retrieve information across sessions, tools, and agent boundaries.
How agents detect failure and decide whether to retry, escalate, skip, or fail gracefully.
The architecture for deciding when agents act autonomously and when they pause for human review.
The input and output filters that prevent agents from receiving or producing content they should not.
Systematic measurement of whether agents produce the right outputs at the right quality level.
Strategies for fitting the right information into the finite context an agent can process.
Connecting agents to external knowledge so they can retrieve facts rather than hallucinate them.
Seven patterns for complex, high-scale, high-stakes AI systems operating at the frontier of what production AI can do today.
Agents triggered by events in your systems rather than by direct user prompts.
Architectures that route agent outputs back as inputs to improve the next cycle.
Many simple agents working in parallel on variations of a problem, synthesised into one output.
An agent hierarchy where strategic agents direct tactical agents that direct operational agents.
Agents that propose improvements to their own configuration — with mandatory human approval.
Two agents take opposing positions; a third evaluates the debate and produces a verified conclusion.
Agents tested against progressively harder evaluation sets, with difficulty dynamically adjusted on performance.
The AIDA certification covers all 21 patterns with a focus on production safety. The CAIG and CAIAUD certs go deeper on governance and audit respectively. All start free.