Boards keep asking whether they are moving fast enough on AI and whether they are exposed if they do. An AI risk register answers both. This board-level guide explains what the register covers, why your existing IT risk framework has blind spots, how to score AI risks by reversibility as well as likelihood, and who should own and review it before anything goes live.

Every vendor now claims AI, and standard procurement checklists cannot tell the platforms from the wrappers. This guide gives executives the due diligence questions that matter: where your data travels, whose model sits under the hood, how to test accuracy before signing, and the contract terms that protect you when the market shifts.

Your people are already pasting work into consumer AI tools, with no rule telling them where the line sits. An AI acceptable use policy is the short, readable document that fixes that: which tools, which data, which tasks, and what is never allowed. This guide explains what the policy should cover, how to tier tools and use cases, how to handle privacy and confidentiality, and how to make it a document people actually follow.

Eighteen months after approving the AI budget, the question lands: what did we actually get for it? This framework helps boards and executives measure return on AI - why it resists easy measurement, the full cost beyond the licence line, where value genuinely comes from, how to capture a baseline, the leading and lagging metrics worth watching, and how long a sensible payback really takes.

Every AI feature that reads untrusted content - emails, documents, web pages - can be instructed by that content to ignore its own rules. Prompt injection is not a bug to patch but a structural property of how language models work. This briefing explains the mechanism without hype, the business scenarios where it bites, the difference between direct and indirect attacks, and the layered architecture - input boundaries, least privilege and human approval - that contains the ri

A senior data scientist with no pipelines to work on, or a junior analyst expected to architect an estate - both are expensive lessons in sequencing. This stage-based capability model separates roles from headcount: engineering and architecture first (foundations), analytics and semantic modelling second (meaning), science and AI specialisation third (advanced value) - with the maturity signals for each transition and the honest build-versus-rent economics, including the hybr