Production AI Institute · Independent certification for production AI practice
Verify a credential|Contact
Insights

Reference library for production AI

Long-form, versioned reference documents, guides, and research on deploying AI safely in production environments. Not blog posts. Citable, maintained, authoritative.

Reference·8 min read

What Is a Production AI System?

A precise technical definition of what separates a production AI system from a prototype, a demo, or a proof of concept — and why the distinction matters for safety, governance, and certification.

2026-04-29Read →
Reference·11 min read

The Seven Failure Modes of Production AI Deployments

An analysis of the most common patterns through which AI systems fail in production environments. Each mode is documented with causes, signals, and architectural mitigations.

2026-04-29Read →
Guide·13 min read

Human-in-the-Loop: When, Why, and How to Design Oversight Correctly

Not every AI decision requires a human checkpoint. Designing oversight correctly means knowing when to require it, how to present decisions for effective human judgment, and how to avoid the trap of meaningless compliance.

2026-04-29Read →
Guide·9 min read

How to Write an AI Behaviour Contract

A behaviour contract specifies what an AI system is permitted to do, in what contexts, with what autonomy level, and under what escalation conditions. This is the foundation of the PAI contract system.

2026-04-29Read →
Practitioner Guide·10 min read

Optimising Your Content for AI Discovery

AI systems now surface content in response to queries that used to go to search engines. The rules are different. This guide covers llms.txt, structured data, answer-first writing, and how to measure AI citation.

2026-04-29Read →
Regulatory·12 min read

What the EU AI Act Means for Your Production AI System

The EU AI Act is in force. If your AI system makes consequential decisions about people, classifies or monitors them, or operates in critical infrastructure, you need to understand what it requires.

2026-04-29Read →

PSF Domain Guides

Deep reference guides for all eight Production Safety Framework domains. Aligned with AIDA, CAIS, CPAP, and CPAA examination content. Licensed CC BY 4.0.

PSF-1Input Governance

Prompt injection defence, input validation, intent classification, and sanitisation practices.

PSF-2Output Validation

Schema enforcement, content filtering, hallucination detection, and confidence gating.

PSF-3Data Protection

PII handling, consent chains, data residency, vector store risk, and erasure obligations.

PSF-4Observability & Monitoring

Inference logging, quality scoring, drift detection, and alerting architecture.

PSF-5Deployment Safety

Canary releases, rollback procedures, circuit breakers, and fallback path design.

PSF-6Human Oversight

Autonomy levels, escalation design, blind review sampling, and skill maintenance.

PSF-7Security

Threat modelling, credential management, adversarial robustness, and red-teaming.

PSF-8Vendor Resilience

Abstraction layers, version pinning, fallback providers, and deprecation response planning.

In development

The following reference documents are currently being written and will be published over the coming weeks.

Industry Playbooks × 5 — Finance, Healthcare, Legal, Retail, Manufacturing (in progress)
The Seven Failure Modes of Production AI Deployments
LLMOps: Operating Large Language Models at Scale
Multi-Agent Systems: Coordination, Conflict, and Control
Red-Teaming Your AI System Before Deployment
AI in Regulated Industries: Finance, Healthcare, Legal
Industry Playbooks × 5 (Finance, Healthcare, Legal, Retail, Manufacturing)
The Production AI Deployment Checklist — interactive, mapped to PSF
Glossary of Production AI Terms — definitive, citable
State of Production AI 2026 — annual flagship report