← AI Terminology
Copilot Pattern
The copilot pattern pairs AI suggestions with a human operator who stays in control — AI drafts or proposes, human accepts, edits, or rejects.
GitHub Copilot popularised the name; it now describes a product archetype across domains.
GitHub Copilot popularised the name; it now describes a product archetype across domains.
Why It Matters in AI
Full automation is often unwanted or unsafe; copilots boost productivity while preserving agency. UX (inline gray text, side panels, accept keys) and evaluation (acceptance rate) differ from autonomous agents.
Key Points
| Aspect | Description |
|---|---|
| UX | Inline suggestions, chat sidebars, one-key accept |
| Design | Low-friction reject; high-quality partial suggestions |
| Metrics | Acceptance rate, time-to-complete, edit distance |
| Examples | Coding copilots, writing assistants, IDE agents |
| Vs agent | Agent acts; copilot proposes |
| Human role | Final authority and editor |
Simple Analogy
A navigator reading the map beside the driver — helpful guidance, but the driver still holds the wheel.
Common Usage Examples
- GitHub Copilot / Cursor tab-complete
- Gmail Smart Compose style assists
- CRM copilots drafting emails
- Measure % of suggestions accepted unchanged
Summary
In short: The copilot pattern keeps humans in control while AI proposes next actions — assistive UX rather than full autonomy.