Strategy / Northbound Frontier
Choosing your first AI workflow
Start with a piece of work you can understand, measure and improve.
Look for a defined task
Choose a workflow with a clear beginning, an identifiable output and someone responsible for the result. Summarising a recurring report is easier to scope than transforming an entire department. Talk to the people doing the work before choosing the technology.
Establish a baseline
Record how long the task takes, what good output looks like and how often corrections are needed. Use representative examples, including awkward cases. These observations give you something concrete to compare a pilot against.
Check information and permissions
Identify the information the workflow needs, who may access it and which actions need approval. Start with a scope that the organisation can responsibly support. Better information access is often a prerequisite to a useful AI system.
Make the next decision explicit
Before building, agree how you will decide whether to stop, revise or proceed. Include quality, effort, cost and user feedback. A useful first project produces evidence for the next decision, even when the initial idea changes.