monkragAll notes ↗
← Back to the graph

Give AI context

Context

Supply the purpose, definitions and limits that make a piece of information interpretable.

2 min readStarter note

Context is the surrounding information needed to interpret material sensibly. For an AI task, it includes the question, the intended reader, relevant definitions, source dates and the limits of what the evidence can establish.

More text is not automatically better context. A short, relevant brief can be easier to use than a large folder containing conflicting versions and unexplained shorthand. Prepare the material as you would prepare a brief for a capable new colleague who does not know your business.

A small-business example

A manager asks an AI tool to summarise “open jobs”. The spreadsheet contains both quotations awaiting acceptance and accepted work awaiting completion. Without definitions, the tool may combine them into a single workload figure.

The manager supplies a field guide, identifies the relevant status values and asks for accepted work only. They also specify the report date and request a list of ambiguous entries. The briefing makes the intended interpretation clearer, while the output still needs checking against the records.

Try this

Prepare a one-page brief for a real task. Include the decision, the source list, three essential definitions and one thing the evidence cannot tell you. Ask a colleague unfamiliar with the work to identify what they would still need explained. Resolve those gaps before handing the material to a tool.

Draw on Define the decision, attach the relevant Data dictionary entries and use Human review to check the resulting interpretation.