Insights

Thinking about AI in production.

Practical perspectives on building, evaluating, governing and scaling enterprise AI.

10 insights

Featured insight

Why enterprise RAG fails in production

Agentic AI & RAG · 13 August 2026 · 5 min read

Most RAG systems fail because they treat retrieval as a search problem instead of an enterprise architecture problem.

Read the insight

Latest insights

Excerpt of the Production Evidence Architecture figure: a demo proves one path worked once, and production requires a system of evidence.

A working AI demo proves surprisingly little

Why production readiness requires traceable evidence across behaviour, context, authority, assurance and accountability, not simply a successful demonstration.

AI Quality Engineering · 6 October 2026 · 10 min read

Read

Before the pipeline, define the purpose

Enterprise AI can connect data across systems, but capability is not permission. A practical framework for governing what AI may know, infer, retain and act upon.

AI Governance & Assurance · 6 October 2026 · 11 min read

Read

How to evaluate AI agents before release

AI agents need more than output checks: task completion, tool-use accuracy, escalation behaviour, safety, cost, and auditability all need evaluation before production.

AI Quality Engineering · 13 August 2026 · 5 min read

Read

AI governance for listed companies

Listed and regulated enterprises need AI governance that produces evidence, not just policy.

AI Governance · 13 August 2026 · 3 min read

Read

Why AI transformation fails without data readiness

AI transformation breaks down when enterprise data is fragmented, poorly governed, inaccessible, or semantically unclear.

Data & Analytics · 13 August 2026 · 3 min read

Read

From workflow automation to agentic automation

Agentic automation moves beyond rules-based task execution, combining agents, approved tools, business rules, human approvals, monitoring, and governance.

AI Automation · 13 August 2026 · 3 min read

Read

Context engineering for agentic AI systems

An event tells an agent that something happened. Context tells it what that actually means, and without context, agents are fragile.

Agentic AI & RAG · 13 August 2026 · 4 min read

Read

How to stop chunks from losing meaning

Retrieval quality lives or dies at the chunk boundary. Overlap, metadata, and the right chunk size keep meaning intact.

Agentic AI & RAG · 13 August 2026 · 4 min read

Read

The unified intelligence layer

Fragmented data, disconnected apps, and inconsistent metrics stop AI working across the enterprise. A unified intelligence layer is the fabric that fixes it.

Data & Analytics · 13 August 2026 · 5 min read

Read

Need something practical?

Checklists, templates and assessments for enterprise AI teams.

Want to apply these ideas to your AI architecture?

We will work through the decisions these articles describe against your own systems, data and controls.

Discuss Your AI Initiative