I’ve spoken to hundreds of data professionals about AI as a solutions engineer and product manager. When talking about AI, folks are rightfully skeptical – oftentimes demos can be contrived or fragile, failing to meet the realities of their business. Our initial approach to AI was crafted with this in mind, allowing for real world AI use cases in production.
Today we’re excited to share the next evolution. Omni’s AI assistant can now run multi-step analytical workflows end-to-end. Like an analyst, it takes an open-ended prompt (“How’s my sales org doing?”) and reasons through the problem, runs analyses, and returns a clear story with the evidence to back it up. The reasoning process is always visible. Before running anything, our AI assistant outlines its task list, and then checks off each step as it finishes. The result is a more transparent, trustworthy, and debuggable process.
This multi-step querying, powered by our new architecture for agentic AI, is live for customers with AI chat enabled — including customers using Omni to embed analytics into their own product.
We’ve heard concerns about similar AI tools going off the rails when asked complex questions. Our AI assistant is grounded in Omni’s built-in semantic layer, so it stays consistent with your business logic, uses the right metrics and joins, respects permissions, and speaks your language. You don’t need to translate business problems into SQL. Just describe what you’re trying to solve, and our AI assistant will take it from there.
What can you do with it? #
Improve sales conversion #
Prompt: “Where's our biggest drop in pipeline conversion? Why is that the case?”
In this video, my teammate Conner gives an overview of Omni’s agentic AI, walking through an example question about sales conversion. You’ll see that our AI analyst creates a task list, runs multiple queries, creates new metrics, and can continue the conversation with you.
Understand product usage #
Prompt: “What features or product areas should I prioritize writing documentation for?”
In this example, Conner walks through a question around product usage and engagement. He shows how our AI analyzed metrics across different datasets, interpreted them independently, and returned a well-reasoned, thorough answer to a difficult question.
Prioritize marketing campaigns based on performance #
Prompt: “Which kinds of campaigns and content should we prioritize next year?”
In this example, Conner shows how our AI created marketing performance metrics and pulled data from multiple datasets to answer the question.
How does it work? #
You can think of our AI orchestrating Omni’s various querying tools so you don’t have to.
With multi-step querying, our AI assistant runs through a “coordinator” mechanism to:
Reason through an analysis plan based on the prompt
Decide what actions to take for each step of the analysis plan (generate a query, create a calculation, use period-over-period functionality, etc.)
Execute queries against your semantic model leveraging existing permission structures and governance
Inspect the results to make sure they are right, and try something else if they aren’t
Synthesize the results into an actionable summary
Omni’s built-in semantic layer prevents the agent from going off the rails. This represents a fundamental shift in how you work with data. In the traditional world of BI, you had to learn the tool. With this new AI-native workflow, the tool learns your business.
Because the agent operates on top of your semantic layer, it already speaks your language, understanding exactly what 'Revenue' or 'Churn' means to your organization. It takes into account:
Security permissions and access controls (i.e. the user will only see data they’re allowed to)
Centralized metric definitions (e.g. a golden definition of ‘Revenue’)
Business context specifically defined for AI (e.g. when someone says ‘Closed Lost’, this means
is_closed= TRUE andis_won= FALSE)
Altogether, this gives you a companion that understands your business, respects security guardrails, and provides repeatable and actionable insights.
Focus on outcomes #
We are moving toward a world where natural language is the primary interface for AI-enabled analytics. By grounding this experience in your semantic layer, Omni’s AI assistant learns how your business works and actively helps you solve problems. It allows everyone in the organization to access governed, trustworthy answers with a simple chat interface.
We’d love to show you this functionality on your own data — please reach out for a demo. For any customers already using Omni, this is all available already in the same AI assistant ( 👋 Blobby) you use every day, including embedded analytics. Please continue to share your feedback over Slack or connect with me directly at jack at omni.co. If you’d like to see what we’re building next, be sure to check out our weekly engineering demos.