Uvi Secures $7.5 Million in Seed Funding to Enhance Context for AI Agents
Agentic artificial intelligence startup Uvi has announced the successful closure of a $7.5 million seed funding round aimed at addressing the costly challenges enterprises face regarding business context. These issues reportedly drain millions of dollars each year from organizations that utilize autonomous AI agents.
The funding round was led by CT Innovations, joined by participating investors Cardumen, Mana VC, and 91 Ventures.
Newly deployed AI agents lack any knowledge about the companies that employ them. Consequently, these agents must undergo a repetitive learning process to understand the company’s business processes and operations from scratch, which leads to substantial costs in computing resources.
The importance of business context cannot be overstated for agentic operations. For AI agents to work autonomously without continuous human oversight, they must possess a comprehensive understanding of the operational intricacies of the organization. This includes familiarity with historical decisions and company procedures. Until now, the only method for agents to acquire this context has been to relearn it with every deployment, incurring fees from model providers for each token utilized in this learning cycle.
Uvi, officially known as Naboo Research Ltd., claims that enterprises are losing significant amounts of money due to this inefficiency. According to their benchmark tests, internally developed AI agents cost the average company approximately eight times more than necessary because of repeated context gathering. This inefficiency also slows down operational speed, as Uvi co-founder and CEO Gilad Salinger notes.
Salinger shared insights with SiliconANGLE, emphasizing that as AI adoption grows, this problem is becoming more pronounced. Initially, companies prioritized the accuracy of AI agent responses and the security of their actions. However, the focus has shifted to cost, token usage, and speed, where these agents continue to encounter challenges.
The Importance of Contextual Understanding
To address these issues, Uvi has developed what Salinger describes as a “context reasoning layer,” which is owned by each individual company. Instead of requiring every new agent to relearn the company’s operations, they can access this unified institutional memory available to all AI tools within the organization’s IT ecosystem.
Uvi facilitates this context by integrating disparate enterprise systems such as Slack, Jira, Confluence, and Salesforce, alongside data from code repositories and internal meeting minutes. This comprehensive mapping creates a complete picture of company dynamics, including the rationale behind decisions and their impacts.
The context reasoning layer operates on top of a knowledge graph that serves as an underlying infrastructure for every deployed AI agent. It employs entity mapping to comprehend the organization’s engineering stack and track dependencies among various assets.
Uvi also examines historical context, analyzing actions such as decisions made in Slack that were subsequently reversed in document comments. This allows the reconstruction of the reasoning processes of AI agents. Furthermore, Uvi has built a custom semantic layer for the knowledge graph, which helps AI agents navigate company-specific jargon for accurate term interpretation.
Significantly, Uvi’s context reasoning layer is designed with security in mind. Acknowledging that knowledge graphs often falter in managing user permissions, the company has implemented an access control model that governs permissions at both application and administrative levels, along with data tagging.
Uvi’s benchmarks indicate substantial cost benefits for companies that adopt its context reasoning layer. While agents operating without this layer can cost between $0.75 and $1.41 per response, and sometimes exceed $2, costs plummet to an average of just 10 to 16 cents per response when utilizing Uvi’s system.
The differences are even more pronounced with open-ended questions, which often see agents trapped in repetitive learning cycles. In such cases, Uvi agents are reportedly 46 times more cost-effective than their counterparts that lack this contextual support, while also delivering responses at double the speed.
Uvi has already demonstrated its platform’s effectiveness to several companies, including Monday.com Inc., Lemonade Inc., and Melio. For Monday.com, Uvi enables agents to comprehend relationships among Jira tickets, pull requests, and feature flags. Without this contextual knowledge, agents often yield disjointed information, necessitating manual sorting by employees to extract relevant insights. By tracing the interconnections of tickets, Uvi significantly expedites the resolution process.
This potential for efficiency has led Gwen Cheni from CT Innovations to express confidence in Uvi’s value. She noted that virtually every company she engages with regarding AI agents struggles to scale context building. “Most are still very far away,” she remarked. “Uvi represents a mature solution for a project that companies have already invested in numerous times.” Salinger believes that a context layer will be essential for all AI agents in the future, as efficiency cannot be maintained otherwise.
“In three years, the enterprise AI stack will not be centered around chat interfaces or document searches,” he predicted. “Instead, it will revolve around a persistent context layer that every agent, model, and identity integrates into. Companies that establish this context will be able to deploy new agents as seamlessly as they currently manage continuous integration and development pipelines, while those that do not will continue to incur costs by retraining their models daily.”