Transform Business Apps with AI: How Retool's Low-Code Platform Simplifies Intelligent Integration

Retool Unveils New Tools for AI-Powered Application Development

Originally designed as a platform for creating business applications, Retool has evolved significantly over the past few years. The well-funded startup recently introduced several back-end services, including a new workflow automation feature. Today, Retool is launching a suite of innovative tools that enable users to develop AI-driven applications, notably a hosted vector store that simplifies adding context to large language models (LLMs).

As David Hsu, CEO and co-founder of Retool, shared, many of his clients are exploring the integration of AI into their applications. However, the key value for businesses lies in their ability to analyze internal data effectively. Current methods, such as copying and pasting data for queries, can be limiting and expensive. While few organizations possess the resources to train their own models, they can successfully fine-tune existing models with a modest dataset. Yet, Hsu emphasizes that fine-tuning with comprehensive production data isn’t practical as that data can quickly become outdated.

At present, the cutting-edge approach to integrating custom data into LLMs is through vectorization, which enhances data accessibility for these models. This trend has led major players like Google, Microsoft, DataStax, and MongoDB to introduce their own vector search services recently.

Hsu notes that these offerings share similar capabilities. “The critical challenge isn’t which vector database to select—be it MongoDB or others. It lies in efficiently getting data into the vectorized database and keeping it current. For instance, syncing it with Salesforce enables LLMs to retrieve fresh data when queried.” Hsu argues that this challenge is where Retool's customers currently face the most significant hurdles in building tailored AI applications. To address this, the company has introduced Retool Vectors, a hosted vector storage service that leverages the open-source pgvector extension for Postgres.

Internally, Retool experimented with Intercom’s GPT-powered AI chatbot for customer service tasks. Initially, the bot, equipped with existing business context, managed to resolve about 20% of incoming tickets. However, after integrating a vector database alongside an LLM that incorporated all of Retool’s Salesforce and support data, the ticket resolution rate surged to nearly 60%. Additionally, Retool has vectorized all its sales call transcripts, layering OpenAI’s API on top for querying.

A standout feature of Retool Vectors is its ability to synchronize a business's production database with the vectorized database through Retool's recently launched Workflow service. This synchronization ensures that models always access the most up-to-date information.

In conjunction with the vector storage service, Retool has rolled out various AI-powered actions tailored for common tasks, such as text summarization, classification, and image generation. These features, developed in collaboration with OpenAI, seamlessly integrate into Retool Workflows.

“As we partner with Retool, we’re eager to help businesses harness the power of generative AI,” remarked Brad Lightcap, COO of OpenAI. “Our joint efforts aim to reduce manual labor, enhance knowledge sharing, and deliver new customer-facing capabilities, ensuring that tools like Retool enable companies to implement AI rapidly and safely.”

Keywords: Retool, AI applications, large language models, vector storage, OpenAI, automation, enterprise software, low-code development.