Starburst adds AI frontend to make traditional business intelligence dashboards obsolete

Starburst Data Inc. today introduced artificial intelligence capabilities designed to help business users query and analyze enterprise data using natural language, positioning the service as a transition from traditional business intelligence dashboards to more interactive, real-time decision-making.

AI Data Assistant aims to address what the company describes as a growing disconnect between the speed of business decision-making and the slow processes needed to build dashboards and reports.

“AI is the new BI, and I think this is going to be the future of the industry around analytics. We’re going to be able to answer questions directly through AI,” Starburst CEO Justin Borgman said during the announcement webcast.

Traditional analytical workflows can require time-consuming extract-transform steps that can no longer keep up with business needs, Borgman said. “Businesses are moving to real-time. Questions change faster than dashboards can be built, and leaders need answers, not queues of tickets,” he said.

This announcement is part of a broader industry trend of applying generative AI interfaces to structured enterprise data. Borgman argued that structured data will be the next major frontier for enterprise AI, citing its role as a system of record for enterprise operations.

“The effectiveness of AI is determined by the data it can access and understand,” he said. “Most organizations don’t have an AI problem; they have a data problem.”

Place your data in place

Aida works across distributed environments without the need for data movement or centralization. This is an important architectural difference from many existing analytics and AI approaches. Starburst’s platform is built on the open source Apache Trino engine to connect to the right data across cloud storage, data lakes, warehouses, and operational systems, and consistently applies governance and metadata across sources.

The company says the assistant uses an inference framework that goes beyond simple text-to-query translation. Instead of generating a single query, it evaluates available data, constructs a query, and iterates through intermediate steps to generate what the company describes as contextual answers.

“This is much more than just a chat,” said Matt Fuller, co-founder and VP of AI. “We reason about what the question means. We identify datasets that can answer the question and ultimately provide the answer.”

Starburst demonstrated that Aida analyzes historical F1 race data and generates queries, visualizations, and explanations based on natural language prompts. In another example using financial data, the assistant created summaries, charts, and executive-level summary information from general ledger data.

Fuller said one of the main limitations of existing BI tools is user friction. He cited internal data suggesting that 41% of customers spend more than four months building a dashboard, and 72% of users regularly circumvent dashboards by exporting to Excel, adding that “67% of businesses are not very confident in their in-app analytics services.”

Aida seeks to address these issues by mapping business terminology to underlying data structures in what Starburst calls a “context layer.” This allows users to ask questions in business language while the system interprets the technical schema.

Choosing an LLM

The platform also supports multiple large-scale language models, allowing customers to choose a model based on cost, performance, or compliance requirements. “Starburst can be connected to the model of your choice, allowing it to fit into your ecosystem,” Fuller said.

Cost control is also a focus. Fuller said 84% of enterprises report AI costs eating into their gross margins by 6% or more, highlighting the need for flexible deployment options, including on-premises and hybrid environments.

While Starburst positions Aida as a complement to existing BI tools rather than a complete replacement, executives acknowledged that its long-term trajectory could result in less reliance on dashboards.

“We are not proposing that dashboards disappear completely, but rather that they become a more specialized and curated set of dashboards,” Borgman said. “We can now use this conversational interface for those extra questions.”

The company outlined a roadmap that includes integrations with enterprise applications such as Slack, Jira, and GitHub, as well as governance features to control AI output and enforce data policies.

“Reports are now becoming conversations, and static BI is being replaced by interactive data,” says Borgman.

Image: Pixabay