GitHub’s 24 September article argues that chat is not always the best interface for working with AI, particularly when a task benefits from a visible, structured surface.
The article uses Copilot canvases as an example: purpose-built interfaces that can work with an agent while giving people direct controls, visible state and a clearer way to interact with the task.
The broader idea is useful beyond developer tooling. Once the job is known, the interface can be shaped around the work instead of around a blank prompt box.
Why it matters
Why it matters: AI may become more useful to small teams when it is embedded in focused workflows, not when every process becomes another conversation. A purpose-built surface can make repeatable work easier to see, steer and review.
A practical rule is to use conversation where exploration helps, and structured interfaces where the work is repeatable, stateful or needs clear human review.
Source: GitHub Blog. This page contains Aegean Astraea’s own summary and commentary; the original article is not republished here.
Read the original source ↗