Most organisations approach AI backwards. They start with the technology — a new model, a shiny tool, a vendor demo — and work backwards to find a use case. The result is a portfolio of proofs-of-concept that never make it to production.
A real AI strategy starts with outcomes, not tools.
What does your organisation actually need?
Before evaluating any AI technology, you need honest answers to three questions:
- Where are our biggest friction points? These are the processes where errors are common, speed is critical, or skilled people spend time on repetitive work.
- What decisions do we make repeatedly? AI is most valuable when applied to decisions that are made at scale with consistent inputs.
- Where does quality matter most to customers? AI can improve consistency, but inconsistency in the wrong place destroys trust.
The answers will tell you where AI can have genuine impact — not where it sounds impressive.
The build vs. buy question
Most organisations do not need to build their own models. Foundation models from leading providers are powerful, cost-effective, and improving rapidly. The real differentiator is not the model — it is the data, the workflow integration, and the human oversight layer.
Build your own when:
- You have proprietary data that would give a model a genuine competitive edge
- The use case requires privacy guarantees that hosted APIs cannot provide
- You have the engineering capacity to maintain and retrain models over time
In every other case, start with an API.
Governance is not optional
The organisations that will build durable AI capability are not the ones that move fastest — they are the ones that move deliberately. That means establishing clear ownership for AI decisions, building evaluation frameworks before deployment, and treating AI incidents with the same seriousness as security incidents.
Getting started with AI strategy is not about picking tools. It is about asking the right questions, in the right order, with the right people in the room.
