What Is an Entity in a Chatbot and How to Use It Well

If the intent answers what the user wants, the entity answers about what. When someone types "I'd like to book a table for 4 on Friday at 8," the chatbot mustn't just understand that the intent is to book — it has to capture the exact details (4 people, Friday, 8:00 PM) to act on them. Those details are the entities. Understanding what an entity is in a chatbot is key to going from a bot that merely "chats" to one that actually gets things done.

A simple definition of entity

An entity is a concrete, relevant piece of information the bot extracts from the user's message. While the intent is the overall goal, entities are the parameters that goal needs in order to be completed.

Think of an intent as a form and of entities as the fields to fill in. For book_table, the entities would be: number of people, date, time, and maybe preferences (patio, indoors).

The most common entity types

Not all entities are the same. These are the categories you'll see again and again:

• System (predefined) entities: dates, times, numbers, currencies, emails, phone numbers. Almost every platform recognizes them with no setup.
• Custom entities: specific to your business, like product names, plans, branches, or service categories.
• List or synonym entities: group variants that mean the same thing. "tee," "t-shirt," and "shirt" can map to a single shirt value.
• Composite or pattern entities: shapes like an order number (#4821) or a license plate, recognized by their format.

A full example

Customer message:

"Hi, I want to change my order #7788 from size M to size L"

The bot processes it like this:

• Intent: modify_order
• Entities:
  • order number → #7788
  • current size → M
  • new size → L

With the intent it knows what to do, and with the entities it knows with what data. Without the entities, it would have to ask for each detail separately, dragging out the conversation and increasing drop-off.

Why entities make the difference

A bot that extracts entities well achieves three valuable things:

1. Fewer back-and-forth questions: it captures several details from a single sentence.
2. Real actions: it can look up an order, book an appointment, or generate a quote because it has the exact parameters.
3. Personalization: it uses the customer's name, product, or city to respond relevantly.

Slot filling: completing the missing pieces

When the user doesn't provide every detail, slot filling kicks in. If they say "I want to book for Friday" but don't mention how many people, the bot detects that entity is missing and asks only for it: "For how many people?" That way it completes the form step by step without repeating what it already knows.

Common mistakes with entities

• Not defining synonyms: if your customers say "cell" but you only trained "cellphone," the bot misses the entity.
• Confusing entity with intent: "size L" is an entity, not a standalone intent.
• Skipping validation: capturing "February 31" as a date without checking it exists.
• Not handling ambiguity: if the customer says "tomorrow," the bot must resolve which concrete date that is based on the time zone.
• Overloading with entities: asking for too many details at once overwhelms and drives people away.

Entities in the generative AI era

With classic chatbots, each entity had to be declared and trained by hand. Modern language models extract entities far more flexibly: they understand "the day after tomorrow in the afternoon" or "the cheapest plan" without your defining every variant. Even so, still thinking in terms of entities gives you control: it ensures critical data (amount, ID, date) is captured in a structured way and can be used in an automation or integration.

On platforms like Omnifox, an AI agent can extract entities from the conversation and pass them straight into a workflow — create a contact, update an order, book an appointment — without a human having to copy and paste anything.

Best practices for 2026

• Identify what data each action needs before designing the bot, and turn it into entities.
• Enrich custom entities with synonyms and regional variants.
• Always validate and normalize: real dates, in-range numbers, correct formats.
• Use slot filling to ask only for what's missing, never for what you already have.
• Review real conversations to discover new values you should recognize.

Conclusion

Entities are the concrete data that turns a conversation into an action: the date of a booking, an order number, a product name. Together with the intent, they form the backbone of any chatbot that truly resolves things. Designing them well — with synonyms, validation, and smart capture — is what lets your customers finish the task instead of giving up. If you want to build agents that capture the right data and act on it, you can try Omnifox and connect them to your workflows in just a few steps.