RelationalAI beefs up its reasoning capabilities to enhance AI agent decision-making

Enterprise decision intelligence startup RelationalAI Inc. is advancing its capabilities for Snowflake Inc.’s AI Data Cloud platform.

At Snowflake Summit 2026 today, it announced a series of updates that aim to provide artificial intelligence agents with the context, reasoning and post-training they need to autonomously make smarter business decisions.

The company said its goal with today’s updates is to close the “AI value gap,” where many businesses struggle to expand on the capabilities of generative AI. The new features in Rel, the company’s flagship agentic decision intelligence system, will help Snowflake customers enhance the capabilities of “decision agents” in areas such as supply chain management, network operations, resource allocation and pricing strategy.

Headlining today’s announcement is the launch of the new Rel App, which is a centralized tool for mapping out a shared and fully governed semantic representation of each customer’s entire business. The app is designed to map everything from a business’s concepts to its internal relationships and operational rules.

Using Rel App, domain experts will be able to explore their company’s data, follow connections and ask questions in their natural language, and reason through decisions. Each interaction will be grounded in the company’s data stored within the Snowflake AI Data Cloud, the company said.

In addition, RelationalAI debuted a key pair of new agent skills it’s calling predictive and prescriptive reasoners. Available starting today, the prescriptive reasoner is designed to handle multidomain reasoning and optimization problems. It combines large language model reasoning with graph math to enhance AI agent’s reasoning accuracy while also reducing computational costs, the company said.

As for the predictive reasoner, also available from today, this leverages graph-based neural networks inside Snowflake to try and forecast business outcomes such as asset failures and customer churn. That means agents can more easily move from making predictions to recommending actions and executing them on behalf of customers.

The new offerings are targeted at a major bottleneck in the broader AI data landscape. As Dave Vellante and George Gilbert of TheCUBE Research noted in their recent Breaking Analysis, platforms such as Snowflake Horizon excel at organizing governance metadata. It acts as a catalog that defines what data means and who can access it.

But enterprises that want a true “system of intelligence” need more than this. By mapping live, executable business logic and relationships into an enterprise ontology, RelationalAI is positioning itself as a critical piece of the enterprise context layer.

“The key distinction is catalog versus intelligence. A catalog gives you definitions. It tells you what a metric means, where data lives, who owns it and what policy applies. That’s necessary, but it’s not enough,” Vellante and Gilbert wrote. “A system of intelligence goes further. It begins to model the business process logic itself – not just the nouns, but the verbs. When business rules, relationships and actions become live, governed and eventually executable, the enterprise moves from metadata to intelligence.”

To support its positioning as a system of intelligence, RelationalAI has become a launch partner of the new Open Semantic Exchange initiative, which makes it simpler for companies to port their business ontologies from third-party platforms into Snowflake, without needing to rebuild anything.

In another update, the startup said its platform now powers the conversational decision intelligence capabilities within Snowflake CoWork, which is a platform that aims to enhance collaboration between human workers and AI agents. With this integration, humans will be able to ask ad-hoc questions about their business in natural language and receive insights derived from RelationalAI’s reasoning system.

Finally, RelationalAI said it’s launching a new “push-button” post-training capability for AI agents in private preview. With this, enterprise users will be able to fine-tune open source LLMs using their own Snowflake data and semantic frameworks. This will support the creation of specialized AI agents that are familiar with the company’s unique terminology and business logic, the startup said.

RelationalAI founder and Chief Executive Molham Aref said that AI agents are similar to humans in that it’s difficult for them to make good business decisions without extensive knowledge of the organization. “With these capabilities running natively in the Snowflake AI Data Cloud, customers can close the AI value gap by giving their agents the context, tools, and post-training they need to take the best possible action in the face of uncertainty,” he said.

Image: RelationalAI