Function calling is a feature that lets an AI model output structured requests to run specific software functions—like checking inventory or sending an email—instead of just replying with text. The app executes the function and returns results to the model.
Why it matters for commerce
Without function calling, an AI can only describe what it would do—like drafting a refund email. With it, the model can actually check order status, apply a discount, or update a CRM record by calling your existing APIs. This is the bridge between a chatbot and an AI agent that gets work done.
How it actually works
You give the model a list of available functions with names, descriptions, and expected parameters. The model reads the user's request, decides if a function applies, and returns a structured JSON call rather than free text. Your code then runs that function and feeds the result back to the model to continue the conversation.
Where it fits with other protocols
Function calling is the base skill; protocols like MCP standardize how models discover and connect to many tools across systems. Think of function calling as the verb and MCP-style protocols as the shared grammar that lets different vendors' tools plug in consistently.
The leadership takeaway
If you're evaluating AI vendors, ask specifically what functions their agent can call in your stack—inventory, billing, support tickets—not just what it can 'understand.' The value is in verified actions, not eloquent responses.
Frequently asked
Is function calling the same as an AI agent?
No. Function calling is one capability an agent uses; an agent also needs planning, memory, and decision logic to chain multiple calls together.
Does the AI model execute the code itself?
No, the model only generates the request to call a function; your application's code actually runs it and returns the result.
Which AI providers support function calling?
Most major model providers, including OpenAI, Anthropic, and Google, support some form of function or tool calling in their APIs.