Stardog Launches Voicebox, a Conversation Layer Driven by LLM, Simplifying Enterprise Data Access

Stardog, a startup company, announced today the launch of a Voicebox, an LLM-driven conversational layer designed to simplify access to business insights.

This solution will be offered as part of Stardog's flagship platform, allowing users to ask questions in plain language and receive answers based on enterprise data without any technical skills.

This move marks the latest effort to simplify data handling by leveraging LLM, joining the ranks of many other data ecosystem participants such as Kinetica, Databricks, and Dremio.

Kendall Clark, co-founder and CEO of Stardog, said in a statement, "It cannot be overstated how much this solution impacts competitiveness and profitability, as widespread access to relevant data has long been one of the biggest obstacles to getting work done. Self-service analytics is no longer limited to programming-savvy technical professionals."

How does Voicebox work?

Essentially, Voicebox works similarly to other natural language processing (NLP) bots, where users simply type in query questions to get answers.

However, the AI layer not only retrieves information from the client platform's raw data but also interacts with Stardog's enterprise knowledge graph.

The enterprise knowledge graph connects all data sources within a company, enriches the information from these sources with relevant context, and creates knowledge that is understandable by both humans and machines. All of this is aimed at providing answers based on timely, trustworthy, and accurate enterprise data through Voicebox.

Clark said, "It achieves this by receiving natural language and using our models to transform human intent into structured graph queries executed by Stardog."

This solution eliminates the task of writing queries for the knowledge graph platform, providing "data democratization" for all knowledge workers and delivering answers without illusions.

Text engineering and proxy technology for Voicebox

It is worth noting that Stardog also uses cutting-edge text engineering and proxy technology to summarize patterns, perform data integrity checks, and ensure the security, trustworthiness, and contextual relevance of user inputs for queries.

Clark further stated, "Stardog's knowledge graph platform also includes other services such as entity resolution, hybrid AI reasoning, and federated graph stream processing, and Voicebox opens up this backend platform to everyone, regardless of their technical skills."

Currently, the company uses a fine-tuned collection of models based on two open-source projects and trains them with data from crowdsourcing projects, research, and synthetic datasets.

It will also add a self-hosted LLM to provide greater business flexibility and create more competitive products for customers. However, the timeline for this stage is not yet clear.

Plans for Voicebox

Although Clark did not disclose the names of the companies using this new conversational AI layer, he confirmed that early access has been provided to dozens of existing and new customers, including companies in the manufacturing and pharmaceutical industries.

He said, "During the project, we regularly engage with them to understand how they want to use LLM in their data projects, see the benefits they have observed, and what they are seeking. Most of our clients and growth projects focus on risk management and compliance in the financial services sector, drug discovery and supply chain management in the pharmaceutical sector, and Product360 and smart factories in the manufacturing sector."

Currently, the company is working on introducing SMS and WhatsApp support for Voicebox, ensuring that the question-and-answer functionality is fully integrated into users' digital workflows. Additionally, it is also considering the possibility of adding support for voice prompts, although there is no public timeline yet.

Since its launch in 2015, Stardog has raised over $23 million in funding and attracted clients such as Boehringer Ingelheim, Schneider Electric, NASA, and the Department of Defense to use its enterprise knowledge graph. According to Forrester's data, the platform can provide a 320% return on investment and over $9.86 million in total revenue within three years.

LLM helps address data challenges

Although Stardog has the advantage of a knowledge graph, it is not the only company simplifying data access through LLM.

In recent months, many companies have simplified different aspects of data processing through generative AI. Kinetica introduced ChatGPT integration for querying data, followed by its own LLM. Snowflake launched Document AI for searching unstructured data, and Databricks introduced LakehouseIQ as a generative AI "knowledge engine" that allows anyone to search, understand, and query internal enterprise data simply by asking questions.

Informatica also launched Claire GPT, which helps users discover, interact with, and manage their data assets through language prompts.