Bot Intents

Every time a client sends a message, the bot classifies it into an intent -- that is, it determines what the client wants to achieve. This classification is the first step in the pipeline and defines how the bot will process and respond to the message.

How classification works

The bot uses artificial intelligence to analyze the client's message along with the recent conversation history (the last 5 messages). The result is:

• An intent (one of 8 categories)
• A confidence level (from 0 to 1)

If the confidence level is below 0.70 (70%), the bot does not attempt to respond and automatically escalates the conversation to a human agent. This prevents incorrect or out-of-context responses.

The 8 intents

1. Property search (search_property)

The client is looking for a property to buy or rent.

Example messages:

• "I'm looking for a 2-bedroom apartment in Escazu"
• "Do you have houses for rent in Santa Ana?"
• "I need a lot of at least 500m2"
• "I'm looking for something furnished that accepts pets"

What the bot does:

1. Extracts search filters (location, price, type, bedrooms, pets, furnished, etc.)
2. Runs a hybrid search on your inventory
3. Presents up to 3 properties with details and links to each property page
4. If there are related projects, mentions them briefly
5. Asks 1 follow-up question to refine the search

If it finds no properties after 3 consecutive attempts, it automatically escalates to an agent.

2. Project search (search_project)

The client is looking for information about development projects (condos, residential communities, plazas, towers).

Example messages:

• "What projects do you have?"
• "I'm interested in Torres del Valle"
• "I'm looking for a condo with a pool"
• "Do you have new pre-sale projects?"

What the bot does:

1. Searches the project catalog using semantic search
2. Presents up to 3 projects with key information: location, price range, construction status, amenities, and project link
3. Offers to show available units within the projects

3. Property details (property_details)

The client asks about details of a property already mentioned in the conversation.

Example messages:

• "Does it have parking?"
• "How big is the lot?"
• "Is it furnished?"
• "How many bathrooms does it have?"

What the bot does:

1. Identifies the property in question based on conversation history
2. Responds with the information available in the database
3. If it doesn't have the requested information, suggests contacting an agent

4. Schedule visit (schedule_visit)

The client wants to see a property in person or schedule an appointment.

Example messages:

• "Can I visit it tomorrow?"
• "I want to see the property"
• "When can we schedule a visit?"
• "I'd like to see the project"

What the bot does:

1. Checks if the conversation has an assigned agent
2. Queries the agent's availability via Google Calendar
3. Offers up to 3 available time slots with WhatsApp interactive buttons
4. If no agent is assigned or no slots are available, escalates to a human agent

See Visit Scheduling for more details about this flow.

5. General question (general_question)

The client has a question not directly related to property search.

Example messages:

• "What are your business hours?"
• "What documents do I need to buy?"
• "Where are you located?"
• "How does the buying process work?"

What the bot does:

1. Searches for the answer in the organization context (the information you configured in Settings > Bot)
2. If it has relevant information, responds directly
3. If it doesn't have the information, suggests contacting an agent

6. Financing information (financing_info)

The client asks about financing options, mortgage loans, or payments.

Example messages:

• "Do you accept financing?"
• "What is the down payment?"
• "Do you offer payment plans?"
• "Do you work with any banks?"

What the bot does:

1. Searches for financing information in the organization context
2. Responds with available information
3. If it doesn't have specific details, suggests contacting an agent for personalized advice

7. Greeting (greeting)

The client sends a greeting or introductory message.

Example messages:

• "Hello"
• "Good morning"
• "Hi there"
• "Hey"

What the bot does:

1. Responds with a friendly greeting
2. Introduces itself as your agency's assistant (using the organization name)
3. Asks how it can help

8. Escalation (escalate)

The client needs or requests human attention, or the bot cannot classify the message with confidence.

Example messages:

• "I want to talk to someone"
• "The bot isn't helping"
• "I need to speak with an agent"
• "I'm frustrated, I need real help"

What the bot does:

1. Informs the client that they will be connected with an agent
2. Disables the bot for that conversation
3. Tags the conversation as "needs attention"
4. Attempts to assign the conversation to an available agent

See Escalation for more details.

Intent summary

Next steps

• Knowledge base -- How the search powering property and project intents works
• Visit scheduling -- Details of the scheduling flow
• Escalation -- What happens when the bot transfers to a human