Definition
Tool use gives AI systems capabilities beyond text generation. A model can decide to call a tool, pass arguments, read the result, and continue the workflow.
How it works
Developers expose tools with descriptions and input schemas. The model selects a tool when needed, the system executes it, and the model uses the returned data to proceed.
Why it matters at work
Tool use is what makes AI agents operational. It also creates permission, security, audit, and rollback questions that teams must govern.
Workplace example
An AI assistant checks calendar availability and drafts meeting options, but cannot send the invite until the user approves it.
Frequently Asked Questions
Is tool use risky?
It can be. Risk depends on tool permissions and task impact. Read-only tools are lower risk than tools that send messages, move money, delete data, or update customer records.