Introduction

Gleen - Generative AI for customer success

Gleen AI is a highly accurate generative AI system that learns everything about your company and products, automatically converses with your customers, and can take actions on their behalf.

Gleen AI

Key Features

Extremely High Accuracy & Relevance - Gleen AI ensures the accuracy and relevance of all generative responses to your customers. Its answers also automatically improve over time.

More Actions - Gleen Action Bots can seamlessly execute almost any action on a customer's behalf, all via generative customer conversations.

Less Set-Up & Maintenance - Gleen AI's ability to ingest imperfect knowledge from multiple sources results in lightning-fast implementation. And maintaining Gleen AI is easy.

Architecture

In building our bot, we harness the power of some of the most sophisticated Large Language Models (LLMs) to ensure accurate and high-quality responses to user queries. Here's a glimpse into the LLMs that play a crucial role in shaping our bot's capabilities:

  1. GPT-3 (Generative Pre-trained Transformer 3)

    • Originated from OpenAI.

    • Recognized for its capability to generate human-like text over a broad spectrum of subjects, thanks to its billions of parameters.

  2. GPT-4 (Generative Pre-trained Transformer 4)

    • Another groundbreaking innovation by OpenAI.

    • It's an evolution of GPT-3, boasting more parameters and refined capabilities. It aims to provide even more coherent and precise outputs.

  3. Cohere

    • The model is known for its excellent rerank algorithms.

  4. Anthropic

    • This model stands out for its large context length of 100k tokens.

    • It ensures interactions that are not only user-friendly but also ethically and contextually considerate.

  5. Llama2

    • It's an open-source model developed by Meta.

    • We have hosted this model on Azure and AWS and fine-tuned it to cater to the customer needs.

Leveraging these LLMs allows our bot to navigate an array of topics and provide responses that are both accurate and contextually relevant.

Last updated