There is a problem every mid-sized and large company has, and almost nobody talks about it honestly. The company has data. It has dashboards. It even has a few people whose entire job is to manage that data. And yet when a department head needs a real answer to a real question, the process typically looks like this: open a ticket, wait two days, receive a spreadsheet, and realise the numbers don't quite match what was discussed in last month's meeting.
Databricks has launched Genie One to attack that problem directly. Genie One is an AI assistant that any employee can talk to in plain English, through Slack, Microsoft Teams, or a mobile app, and get answers from the company's actual data. No technical skills required. No new interface to learn for most people.
The real question with any tool like this is accuracy. AI assistants are well known for giving confident-sounding wrong answers. Databricks' solution to that is what they call Genie Ontology: a system that, before answering anything, reads how the company already defines and uses its own data. It looks at past queries, certified dashboards, and reports, then weighs which sources are most authoritative. The logic is similar to how Google decides which web pages to show first: sources that are widely trusted and frequently used get more weight than obscure ones.
In Databricks' internal testing, this approach got 84.5% of questions right on the first try. General-purpose AI coding tools managed 52.4% at best, and just 25% at worst. That is a significant difference when the question is something like "what was our gross margin in the North region last quarter" and the answer is going into a board presentation.
Genie One also lives where work actually happens. Employees can tag it in a Slack or Teams conversation, ask a question, and get an answer without switching applications. For organisations that already use tools like Microsoft Copilot or other AI assistants, there is a connection option so Genie can work alongside those tools without requiring a full switch.
On pricing, the model is shifting. Starting July 6, 2026, Genie products move to pay-as-you-go billing. Each user keeps a free monthly allowance covering roughly 80 to 100 questions per month. Usage beyond that is billed by consumption, with no per-seat fees. For most employees, that free tier is probably enough for typical use.
The real-world results from early adopters are worth noting. The UK's Automobile Association integrated Genie into Microsoft Teams and saw a 70% efficiency gain in answering routine data queries, freeing their data team for deeper analysis. At Webmotors, a Brazilian automotive marketplace, IT support tickets for data questions dropped by 72% after deployment.
The competitive context matters here. Snowflake has its own natural-language querying product, and Microsoft's Power BI has Copilot built in. The difference Databricks is stressing is that Genie Ontology automatically builds its understanding of your business without someone having to manually set it all up, and that it enforces data permissions automatically, so employees only see data they are authorised to access.
For any organisation already running Databricks as its data platform, this is the most direct path to giving non-technical managers the ability to answer their own questions. For organisations not on Databricks, the competitive pressure from this kind of product will show up in how their existing tools respond over the next 12 months.