Practical AI starts with the task, not the tool

The organisations getting value from AI did not start with a model. They started with a queue of work nobody wanted to do.

20 August 2026 · 4 min read

Every board has now asked what the organisation is doing about AI. The weakest answer is a pilot chosen because a vendor offered one. The strongest is a list of tasks: where skilled people spend hours on unskilled work, where judgements are made inconsistently, where a queue never quite clears.

Start there and the technology choice becomes obvious, and often modest. Triage that reads an inbound referral and routes it. A first draft that a person checks and signs off. A reconciliation that runs overnight instead of on a Friday afternoon.

Governance comes first. Every automated step should be explainable, logged and reversible, with a person accountable for it. That is what makes the second and third use case easy to approve.

And sometimes the honest answer is that a better form, a cleaner process or a simple rule does the job. We would rather tell you that than sell you a model.