AI
Before starting an AI project, confirm four things: a specific repeated task, accessible data, a defined confidence boundary, and a named owner after launch.
In order
'Classify inbound enquiries by service area' is a project; 'use AI in marketing' is a budget line with no completion criterion. Frequency matters too: a task performed twice a month rarely repays the build.
Whatever the task needs (documents, records, history) must be accessible and reasonably consistent. This is where most stalled initiatives actually stopped, and checking it takes days rather than the months a data programme would.
Where does the system stop and hand to a person, what context travels with the handover, and what does the person see. A system that answers everything answers wrongly at the edges, and that is a design choice rather than an accident.
Models change, prompts drift, source documents go stale and providers alter terms. An AI system without a named owner degrades quietly, and the first sign is usually a customer complaint.
No. You need the specific data the first task requires to be accessible and reasonably consistent. That is far smaller than a data programme and can be done project by project.
A repeated task with unstructured input, a human check, and errors that are visible and cheap to correct. Irreversible or regulated decisions are the wrong place to start.
Before starting an AI project, confirm four things: a specific repeated task, accessible data, a defined confidence boundary, and a named owner after launch.
You have read what we think. If you want to know what it means for your case specifically, describe it and we will tell you which parts of ai project checklist actually apply — and which do not.
