Project Stats

Name
Go2Africa
Industry
Travel
Employees
100+
Location
Cape Town, South Africa
Year
2023
Duration
6 Months
Team
1 AI Engineer, 1 Full-Stack Engineer
TechStack
  • LangChain
  • Docker
  • Scrapy
  • Slack API
  • Weaviate
  • PostgreSQL
KeyResults
Go2Africa company logo representing leading safari and travel planning services.

Go2Africa | Transforming Customer Support with AI-Powered Automation

Project Overview

Go2Africa (G2A) is a safari and travel planning company that is currently experiencing rapid growth. They needed to enhance their customer service operations while maintaining their reputation for excellence in client experiences. With a large volume of inquiries, their goal was to streamline support, automate repetitive tasks, and improve response times without sacrificing quality.

We implemented an AI-driven solution to automate Level 1 customer support tasks, including drafting responses to common inquiries and improving internal efficiency through automated data extraction. The project leveraged advanced technologies such as GPT, Agents, and Visual LLM, integrated with G2A’s Microsoft mail servers, internal policies, and knowledge bases.

Goals

1. Automate Customer Support:

Develop an AI-powered system to handle Level 1 inquiries, such as FAQs and common customer queries, reducing reliance on manual interventions.

2. Improve Response Quality and Speed:

Enable the AI bot to draft accurate responses by training it on six months’ worth of emails, blogs, FAQs, and internal documents, ensuring consistent and context-aware communication.

3. Streamline Data Extraction:

Automate the collection and processing of structured information from merchant websites, focusing on accuracy and scalability.

4. Ensure Seamless Integration:

Connect the AI system with G2A’s existing tools, including Outlook and internal CRM systems, for efficient deployment and adoption.

Challenges

1. Handling Complex Customer Queries:

Customer inquiries often involved intricate questions about itineraries, safari options, and travel logistics. The AI needed to understand these nuances and provide contextually accurate responses.

2. Training the AI Bot:

The bot was trained on a diverse set of data sources, including historical emails, training materials, and policy documents. Ensuring the bot could accurately interpret and cite these sources was critical.

3. Overcoming Anti-Bot Blocking Technology:

Scraping merchant and travel websites presented challenges due to advanced anti-bot mechanisms and dynamic content loading via JavaScript.

4. HTML Noise in Scraped Data:

Extracting meaningful data required filtering out extraneous HTML elements and consolidating scattered information across multiple pages.


Solutions

1. AI-Powered Customer Support:

Automated Drafting: Developed a GPT-based bot capable of drafting responses to common queries, significantly reducing response times.

Data Integration: Connected the bot to six months’ worth of email history, blog posts, FAQs, and training materials. This enabled it to provide context-aware answers while citing sources for transparency.

2. Advanced Data Extraction:

Agents and Visual LLM: Used Agents to dynamically navigate complex travel websites and extract structured data. Visual LLMs were employed to process visual elements like menus and schedules.

Custom Algorithms: Designed algorithms to clean HTML noise, reducing unnecessary data and improving the efficiency of GPT calls.

3. Overcoming Anti-Bot Protections:

• Implemented adaptive techniques, such as randomized request headers and timed delays, to bypass anti-bot mechanisms without disrupting website functionality.

4. Streamlined Integration:

• Integrated the AI system with G2A’s Microsoft mail server and CRM tools to centralize operations and streamline workflows.


Results

1. Increased Efficiency in Customer Support:

The AI bot automated 80% of Level 1 customer inquiries, allowing human agents to focus on more complex and high-value tasks.

2. Faster Response Times:

Average response times for Level 1 queries were reduced by 60%, improving customer satisfaction.

3. Enhanced Data Accuracy:

By leveraging Agents and Visual LLM, G2A was able to extract and consolidate accurate data from merchant websites, ensuring reliable travel information for customers.

4. Cost Savings:

Automating repetitive tasks reduced the need for outsourcing, leading to significant cost savings in customer support operations.

5. Scalability:

The AI-driven solution is scalable, providing a foundation for future automation initiatives, including itinerary generation and advanced training tools.

Ready to grow your idea 10x?

MVP, POC, production-grade pipelines, we have built it all

No salespeople, no commitment