# What is AI Copilot?

Also called copilot, inline assistant.

An AI copilot is an assistant embedded inside an application that offers suggestions in place while a person works, with that person accepting, editing, or rejecting each one. The name comes from code completion tools and now covers writing, spreadsheet, design, and support software. The defining property is that the human stays in control of every change.

The copilot pattern trades autonomy for immediacy. Because the suggestion appears exactly where the work is happening, there is no context to re-explain and no output to carry between systems. Because a person confirms each suggestion, mistakes surface at the moment they are made rather than being discovered later inside a finished document or a shipped change.

Adoption is usually the easiest of any pattern in this category, since nothing about how work is organized has to change. Reported gains concentrate in first drafts, boilerplate, and lookups within the application. Gains are much smaller for work whose difficulty lies in deciding what to do rather than in producing the text, the formula, or the code.

Copilot has become a generic product name and is now attached to features that act on their own, which contradicts the original meaning. When comparing products it is worth asking whether a feature suggests and waits, or executes and reports afterwards. Those two behaviors carry very different review requirements and very different consequences when the output is wrong.

A copilot sits inside one application, an assistant is addressed directly and ranges across topics, and an agent carries a task from start to finish. Many products contain all three modes. They are complementary: a person can accept copilot suggestions in the document in front of them while an agent works a queue in the background.

## Key points

- Embedded in an application and suggests in place
- Human accepts, edits, or rejects every suggestion
- The name is now applied to autonomous features too
- Largest gains on drafts, boilerplate, and in-app lookups

## In practice

A solo founder writes a support macro in her help desk. As she types, the copilot proposes the next sentence based on the twenty macros she wrote before, and offers a shorter version of a paragraph she flagged as clumsy. She takes the shorter version, rewrites the closing line herself, and saves. Nothing was sent or changed without her clicking.

## Related terms

- [AI Assistant](/en/glossary/ai-assistant)
- [AI Employee](/en/glossary/ai-employee)
- [Workflow Automation](/en/glossary/workflow-automation)
- [AI Adoption](/en/glossary/ai-adoption)

[Back to the AI Glossary](/en/glossary)