Decagon Customer Support Automation with OpenAI

Decagon uses a combination of OpenAI models to automate high-volume customer support for enterprises and startups, enabling some clients to handle 91% of global support conversations without human intervention. This approach allows businesses to maintain high quality and speed while managing millions of interactions across the entire customer lifecycle.

Multi-Model Architecture for Specialized Tasks

Decagon optimizes performance by assigning specific OpenAI models to different stages of the customer interaction pipeline based on their individual strengths. This architectural flexibility allows the platform to capture complex business logic and create software surface area that was previously impossible before the advent of Large Language Models (LLMs).

Model-Specific Applications

• GPT-3.5 (Fine-tuned): Decagon fine-tuned GPT-3.5 specifically to rewrite customer queries before they enter retrieval-augmented generation (RAG) workflows, achieving higher performance than other model configurations for this specific task.
• GPT-4: This model is utilized for complex decision-making tasks, including the efficient processing of API requests and other intricate operations.
• Broad Model Suite: The platform integrates GPT-3.5, GPT-4, GPT-4o, GPT-4 Turbo, and OpenAI o1-mini to deliver agentic bots capable of managing the full customer lifecycle.

Scalability and Deployment Efficiency

Decagon's infrastructure allows for rapid deployment and high-performance scaling. For new customers, the core infrastructure can be operational within days. This speed is attributed to the combination of OpenAI's models and Decagon's tailored workflows, which minimize latency—a factor CTO Ashwin Sreenivas identifies as having a direct impact on customer satisfaction.

Evaluation and Iteration

Decagon maintains a competitive edge by rapidly evaluating new model releases. The company employs a rigorous evaluation process to integrate new OpenAI models as soon as they are released, ensuring the platform consistently uses the most accurate and efficient versions available.

Future Directions in AI Support

Decagon is currently expanding its capabilities to support a wider range of customer needs, with a primary focus on exploring voice capabilities. The goal is to apply the same level of automation and accuracy to voice-based customer support interactions as they have achieved with text-based automation.