Genie One, Databricks’ new agent colleague, brings AI automation to every part of your business

Big data company Databricks Inc. is getting into the agentic artificial intelligence coworker game with the announcement of a new tool called Genie One that aims to help business teams orchestrate workflows and automate work-related tasks.

The arrival of Genie One expands on the company’s existing Genie suite, but goes far beyond its original conversation analysis capabilities. Rather than simply digging into data for answers, we provide comprehensive support by doing the work for your employees. Infer both structured and unstructured data, including enterprise information that exists outside the Databricks platform.

Genie One was announced at the company’s annual Data + AI Summit, which kicked off today in San Francisco. At the event, the company also announced a new architecture called Lake Transactional/Analytical Processing that consolidates data from operational and analytical workloads into a single data lake.

Co-founder and CEO Ali Ghodshi said Genie One is an effort to help companies overcome frustration with existing AI co-pilot tools that have not lived up to their initial, unconventional promise. AI has had a huge impact on software engineering teams, but only because AI coding tools have had the good fortune of embedding the necessary context within the source code they are working on. But other business workloads don’t have the same luxury. When it comes to sales, marketing, finance, etc., the critical business context needed to automate tasks is highly fragmented and exists, but it can be deeply buried in disparate software platforms, years-old business documents, widely scattered, or even locked inside the heads of employees.

The challenge is that standard AI agents tend to take a holistic view of the business. This means it’s difficult to reliably answer simple business questions, let alone actually automate tasks. In the absence of context, many AI tools simply try to fill in the blanks and risk guessing, which can have disastrous consequences in highly regulated fields such as finance and operations.

With Genie One, you don’t have that problem thanks to the Genie ontology. The Genie ontology is a self-improving context layer that maps the scope of an organization’s knowledge by scanning all business data, documents, content, applications, and even learning from its employees. The Genie ontology gives your new assistant a more thorough understanding of your business. Continuously extract business knowledge from all sources you have permission to access, including Databricks itself and other sources such as connected workplace applications, files, tickets, chat apps, and meetings. This embedded “ground truth” allows Genie One to create answers based on real, trusted business knowledge. This means you can take the right action instead of guessing. As a result, Ghodsi claimed, latency and cost are lower, and accuracy is much higher.

“Most enterprise AI today is just guessing with false confidence, but that’s not enough for business,” he said. “If you’re a CFO and AI can’t tell you why your margins have changed, or you’re a sales leader and can’t find your next upsell, it’s not an AI problem; it’s a context problem. Genie Ontology continuously learns context from data everywhere, so our answers are much faster and our agents are more accurate.”

Rather than inferring and hallucinating over a few fragmented documents, Genie One works by getting the context you need through structured query language queries. But it’s more than just answering questions. It also features a visual interface, including interactive graphs, and allows users to set alerts. Additionally, integration with Model Context Protocol allows you to use third-party software and tools to perform actions within any business workflow.

Along with Genie One, customers will also have access to the first Genie Agent, which is generally available today. Users can transform any conversation they have with Genie, the company’s original AI chatbot, into a “reusable agent” that inherits the original source data, instructions, and behaviors. Employees can use these agents to perform repeatable workflows and improve productivity. Others include the new Genie App Builder, which provides a comprehensive vibe coding environment for business workers to upload context and generate previews of applications connected to that data, fully secured by Databricks Unity Catalog.

Finally, the company updated Genie Code and introduced Genie ZeroOps for data engineers. The first is a tool that helps data engineers plan, create, and execute data engineering and analysis workflows. Added the ability to track progress and review each step of different projects. The latter, on the other hand, is a new background agent designed to autonomously monitor, investigate, and suggest modifications to data pipelines, tables, machine learning models, and more.

Databricks said it will eschew Genie One’s traditional Software-as-a-Service licensing model in favor of simpler pay-as-you-go pricing, where customers pay for the tokens they consume.

Photo: Rob Hoff