AI and the Analytics Operating System
An operating system is only meaningful if it holds under pressure. AI is pressure. Not because it is a threat — but because it exposes everything that was already fragile. Poor data foundations. Undefined ownership. Ungoverned outputs. AI does not create these problems. It inherits them, and amplifies them.
There are two places where AI intersects meaningfully with analytics work 👇
1
Analytics Delivery Vendors now promise you can retire your dashboards and simply chat with your data. What they omit: the semantic models and preparation work required to make that function represents the majority of the effort — the same effort most organisations have been deferring. Poor data foundations constrain a chatbot exactly as they constrain a dashboard. Chat interfaces also remove the shared, validated views that give teams a common reality to work from. Without those, you do not have better analytics. You have faster disagreements.
2
True AI Deployments RAG systems. LLMs applied to internal knowledge. Agentic workflows that act on your data. This is where real transformation lives. These systems are only as good as the data, metadata, and governance underneath them. A RAG system fed poorly labelled documentation will retrieve the wrong answer — confidently. An agentic workflow built on fragile pipelines will automate your chaos, not resolve it.
What AI Actually Requires
Every AI initiative generates a new monitoring requirement. How is the model performing? Has it drifted? Where are the failure modes? These are analytics questions. They demand the same rigour as any other: clear requirements, stable pipelines, validated outputs, explicit ownership.
Before identifying where AI will help, you need data. You need to know which decisions are slow, which processes are error-prone, which parts of the organisation run on instinct rather than information. That diagnostic work is analytics work. It requires the four pillars to already be functioning.
Non-Negotiables Before Any AI Initiative
- No AI deployment without defined data ownership.
- No automation without understanding what is being automated.
- No model in production without a monitoring plan.
Without these, AI does not add capability. It adds surface area for failure.
The Right Use of AI Momentum
The right response to the AI moment is not to chase it — it is to use it as a forcing function.
Let the momentum of AI push you to look hard at how your business actually operates: what processes exist out of habit, what decisions are made slowly because the data is not there, what work consumes energy without creating value.
- Strip out what is useless.
- Automate what is repeatable and well-understood.
- And where your business truly differentiates, where judgement, relationships, and context matter, keep humans close. Those are precisely the areas where automation creates risk, not leverage.
AI is a lens before it is a tool. Use it to see your business more clearly.
The scenarios above are illustrative. The full picture — how AI and data analytics interact across different organisational contexts, maturity levels, and use cases — is covered in the book. For regular updates on how these dynamics play out in practice, join "The Simplicity Stack"