How Intercom is redefining customer support with Claude

Key Concepts:

• AI in customer support
• Large Language Models (LLMs)
• Automation of repetitive tasks
• Efficiency gains for businesses and end-users
• Trustworthiness and accuracy of AI models
• A/B testing for model evaluation
• Multilingual support
• Revenue generation through AI products

1. Introduction: AI Revolution in Customer Support

• Fergal Reid, VP of AI at Intercom, highlights AI as a defining technological revolution.
• Intercom, a customer service software company, is shifting its focus to AI as the future of its business.
• AI is poised to completely transform customer support and customer service.

2. The Role of AI and Humans in Customer Support

• Distinction between tasks suitable for humans (empathy, judgment) and AI (repetitive tasks).
• Human representatives answering the same question repeatedly can become bored and lose concentration.

3. Fin: Intercom's AI-Powered Customer Support Product

• Fin is designed to answer repetitive end-user customer support questions quickly and accurately.
• Fin leverages a business's customer support knowledge base to provide replies.
• The goal is to save time for both the business and the end-user, increasing efficiency.

4. Anthropic and the Importance of Trustworthy AI

• Anthropic is building models with a focus on trustworthiness and accuracy, aligning with Intercom's values.
• Intercom values trustworthy and accurate AI.

5. Claude Sonnet 3.5: A High-Performing Language Model

• Claude Sonnet 3.5 impressed Intercom, leading to production testing.
• Massive A/B test with millions of end-user interactions was conducted to evaluate Claude.
• Claude emerged as the winner of the A/B test, leading to its selection to power Fin.

6. A/B Testing and Model Selection

• Intercom uses A/B testing to evaluate the performance of different language models in real-world scenarios.
• The A/B test involved millions of end-user interactions, providing statistically significant results.
• Claude's superior performance in the A/B test was the deciding factor in its adoption.

7. Multilingual Support and Global Reach

• Modern language models like Claude excel at translation between many languages.
• Fin supports over 45 languages, including Chinese, German, and Japanese.
• Claude handles the translation layer, enabling Fin's global reach.

8. Fin's Success and Revenue Generation

• Fin has been a highly successful product for Intercom.
• Fin generated eight figures in revenue in its first year.
• Anthropic's technology has contributed to Fin's success.

9. Conclusion: The Impact of AI on Customer Support

• AI is revolutionizing customer support by automating repetitive tasks and improving efficiency.
• Trustworthy and accurate language models like Claude are essential for delivering high-quality AI-powered customer support.
• Multilingual support enables businesses to provide customer support to a global audience.
• AI-powered customer support solutions can generate significant revenue for businesses.