Short answer
Slot filling in NLP is the technique of extracting the specific pieces of information an intent needs before it can be acted on, such as an order number, a return reason, or a delivery date.Once a system knows the customer's intent, it usually needs details to resolve it. Those details are the slots. A cancel order intent has slots for the order ID and the cancellation reason. A change address intent has slots for the new address and the order it applies to. If a slot is missing, the system asks a follow-up question until it has what it needs.
Slot filling is what turns a recognized intent into an executable action. Intent tells you what the customer wants. Slots give you enough to actually do it.
Slot filling vs named entity recognition at a glance
| Dimension | Slot filling | Named entity recognition |
|---|---|---|
| Purpose | gather what an intent needs to act | find and label entities in any text |
| Tied to an intent | yes, slots are defined per intent | no, it runs regardless of intent |
| When a value is missing | asks a targeted follow-up | nothing, it only extracts what is present |
| Example output | order ID and reason for a cancellation | dates, names, order numbers |
Aide, the AI platform for customer service, treats slots as part of an intent-scoped automation, not a free-floating form. After classifying intent, an ASOP (an Agentic SOP) defines the slots that intent requires and the conditions for acting on them. Many slots fill themselves from Shopify, WooCommerce, or Salesforce data, so the customer is asked for less.
An automation never acts on a half-filled or low-confidence slot: missing information triggers a question or a handoff, never a guess. And because the team authors the slot logic itself, the people who run the workflow know exactly what the AI needs and why.