My team and I have developed a customer service platform, leveraging the capabilities of OpenAI’s GPT 3.5 and 4, Bing’s custom search, and internal document search functionalities.

Its key benefit is the ability to summarize and cite the top search results.

This AI system makes use of both public and internal search results to provide citations for specific facts. This not only brings more transparency to the generated answers, but also helps improve retrieval accuracy and the overall quality of the response.

We believe in the power of combining the intuitive and transparent UI of web search with the intelligence of large language models. The underlying search and indexing engine provides the necessary context while the interpretation is done by the large language models. To improve the quality of responses, we employ two LLM calls - one to generate multiple queries to retrieve diverse results, and another to generate the final response.

We’ve developed a playground where you can quickly test our system on any website - visit playground.algomo.com.

It’s like integrating perplexity.ai into your customer service. On our full platform at algomo.com, you can use multiple websites and even upload your own documentation from Notion.

We’re gradually developing new features for example allow users to view and edit the generated search queries, integrate other APIs, and AI cobrowsing.

Curious what the community thinks, and what features come to mind when you use this interface.