Customer support · Production

Agoda builds Property AMA Bot to cut property question response time from 8 hours to seconds

The problem

Users had to sift through long property pages to find specific answers, and when they used Agoda's Customer Messaging platform to ask properties directly, only 55% of questions received a response and the average wait was eight hours.

First attempt

The existing Customer Messaging platform allowed users to contact properties directly but fell short because many inquiries went unanswered and responses took an average of eight hours.

Workflow diagram · grounded in source
1
User submits property question
trigger
“The chatbot client acts as the front-facing component, interacting directly with users. It is designed to interpret user queries and connect seamlessly with a bot, ensuring a smooth conversational experience.”
2
Input topic guardrail check
validation
“Input topic guardrail — to check if the user's question is allowed based on our provided topic list”
3
Property data fetched from PAPI
integration
“we connect to Property API (PAPI) to fetch property information that enriches the chatbot's knowledge base”
4
ChatGPT generates answer
ai_action
“the question is parsed and relayed efficiently to ChatGPT for processing with context that is fetched from Agoda services”
5
Instant answer delivered to user
output
“users can now ask the bot directly and receive an instant reply”
6
Response labeling and A/B testing
feedback_loop
“By applying these labels to chatbot interactions, we gain valuable insights that enable developers to refine prompts, adjust data inputs, and optimize algorithms, ultimately improving the chatbot's accuracy and reliability.”
Reported outcome

The Property AMA Bot provides instant answers to property-related questions in seconds, replacing an average eight-hour wait, and is accessible across desktop, mobile web, and the Agoda app.

Reported metrics
Property question response time (pre-bot)average of eight hours
Property question response time (with bot)seconds
Properties with inbox messaging enabled80%
Questions receiving a response from properties55%
Reported stack
ChatGPTLLM APIProperty API (PAPI)AIDEOpenAI
Source
https://medium.com/agoda-engineering/how-we-built-agodas-property-ama-bot-to-simplify-travel-decisions-b861c7ec7ff1
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The Property AMA Bot provides instant answers to property-related questions in seconds, replacing an average eight-hour wait, and is accessible across desktop, mobile web, and the Agoda app.

What tools did this team use?

ChatGPT, LLM API, Property API (PAPI), AIDE, OpenAI.

What results were reported?

Property question response time (pre-bot): average of eight hours; Property question response time (with bot): seconds; Properties with inbox messaging enabled: 80%; Questions receiving a response from properties: 55% (source-reported, not independently verified).

What failed first in this deployment?

The existing Customer Messaging platform allowed users to contact properties directly but fell short because many inquiries went unanswered and responses took an average of eight hours.

How is this customer support AI workflow structured?

User submits property question → Input topic guardrail check → Property data fetched from PAPI → ChatGPT generates answer → Instant answer delivered to user → Response labeling and A/B testing.

Related customer support cases