The promise of an AI agent that can act on your behalf inside a company’s systems is a compelling one. The reality, for most enterprises, is a confident chatbot that can summarize a document but has no memory of why a decision was made last quarter, or which engineer owns the failing service. The gap between promise and production is a layer of context that is structured, historical, and decision-aware. That is the layer Uvi is trying to build.
Founded in 2023 by Gilad Salinger and Dror Wolmer in Tel Aviv, Uvi emerged from stealth in October 2026 with a $7.5 million seed round. The company calls its product an “AI reasoning layer,” a piece of infrastructure designed to sit between an enterprise’s data sources and the AI agents that query them. Its core offering is not another large language model, but a system that continuously ingests information from code repositories, logs, support tickets, and documentation, then organizes it into a live graph of relationships and past decisions [Uvi team page]. The bet is that for an AI agent to be truly useful, it must understand not just what information exists, but the rationale behind previous actions [The Next Web, October 2026].
The architecture of memory
Uvi’s technical wedge is its “Decision Graph,” built by CTO Dror Wolmer. The platform operates as a live ETL (extract, transform, load) pipeline, pulling from disparate enterprise systems to construct a constantly updated map of a company’s operational knowledge. This structure allows an agent to answer questions like “Why did we switch database providers in Q3?” by traversing connections between meeting notes, cost analysis tickets, and subsequent deployment logs, rather than simply retrieving a static document [Uvi team page]. The platform is model-agnostic, supporting agents built on top of Claude, Gemini, and GPT [The Next Web, October 2026].
The commercial claim is efficiency. Uvi reports that its structured context can reduce the cost of answering complex, multi-source queries by eight times compared to alternatives, bringing the cost per answer down to between 10 and 16 cents [SiliconANGLE, October 2026] [TokenPost, October 2026]. In its own testing on multi-source questions, Uvi says its system provided correct answers 88% of the time, versus 75% for baseline methods [Uvi blog, Introducing Uvi]. For engineering and operations teams deploying agents, the value proposition is moving from costly, often incorrect summarization to cheaper, actionable intelligence.
Early traction with Israeli tech leaders
Uvi’s initial market entry appears focused on a cohort of sophisticated, technology-native companies, many based in Israel. The company lists several early customers, including insurance tech firm Lemonade, work management platform monday.com, and small business payments company Melio [The Next Web, October 2026]. A sales job posting also names Global-e and Fiverr as customers [Uvi sales job posting]. While the depth and commercial terms of these engagements are not public, they represent a beachhead in the complex, high-stakes environments where agentic AI could deliver the most value.
The $7.5 million seed round, led by CT Innovations with participation from Cardumen Capital, Mana VC, 91 Ventures, and Gaingels, is earmarked for team growth and U.S. expansion [The Next Web, October 2026]. The company is actively hiring for an Enterprise Sales role and a Head of Customer Success in the United States, signaling a clear intent to move beyond its initial regional footprint and into broader enterprise sales cycles [Uvi careers].
| Founder | Role | Key Background |
|---|---|---|
| Gilad Salinger | Co-founder & CEO | Previously CEO and co-founder of Naboo, an enterprise AI context layer [Forbes, March 2026]. |
| Dror Wolmer | Co-founder & CTO | Built Uvi’s Decision Graph and live ETL infrastructure; previously associated with Naboo [Uvi team page]. |
The crowded context layer
Uvi is entering a space that is rapidly becoming crowded with startups aiming to solve the “context problem” for AI. Competitors include companies like Zep, Mem0, and Modern Relay, each with different architectural approaches to managing memory and knowledge for agents. The risk for Uvi is that its highly structured, decision-centric approach, while powerful, may require more upfront integration work and semantic modeling than lighter-weight alternatives. Enterprises may opt for a simpler, faster-to-deploy tool for initial agent experiments, potentially leaving Uvi’s more comprehensive solution for a later, more mature phase of AI adoption that has not yet materialized at scale.
Uvi’s answer to this is likely its focus on performance and cost. By anchoring its value on measurable outcomes,higher accuracy and an eightfold reduction in query cost,the company is betting that economic pressure will eventually push enterprises toward the most efficient architecture, not just the easiest one to install. The early adoption by tech-forward companies suggests this message is resonating where the pain of inaccurate or expensive agent queries is already felt.
What to watch in the next year
The next twelve months will be critical for Uvi to translate its technical vision and early lighthouse customers into a repeatable commercial motion. Key milestones to watch include the closure of its first major U.S. enterprise deal, the publication of any third-party validation of its performance claims, and the evolution of its partnership strategy with major AI model providers and agent framework companies.
The company’s success hinges on a specific and challenging workflow: the deployment of autonomous or semi-autonomous AI agents in complex engineering and business operations. Today, the standard of care in most organizations is either manual,a human operator digging through Slack, Jira, and Confluence to piece together context,or a basic retrieval-augmented generation (RAG) system that fetches documents but fails to connect decisions across time and systems. Uvi is betting that the next phase of enterprise AI requires a deeper, more structural form of organizational memory. For teams drowning in tribal knowledge and decision debt, that memory could be the difference between an AI assistant that suggests a change and one that understands why the last one failed.
Sources
- [The Next Web, October 2026] Uvi raises $7.5M to give AI agents the company context they lack | https://thenextweb.com/news/uvi-7-5m-seed-ai-agents-company-context
- [SiliconANGLE, October 2026] Uvi raises $7.5M to help AI agents learn business context and run at eight times lower cost | https://siliconangle.com/2026/10/14/uvi-raises-7-5m-help-ai-agents-learn-business-context/
- [TokenPost, October 2026] Uvi Says Its AI System Cuts Answer Costs to 10 to 16 Cents | https://www.tokenpost.com/news/technology/31767
- [Uvi team page] Uvi team page | https://www.uvi.io/team/
- [Uvi blog, Introducing Uvi] Uvi blog post | https://www.uvi.io/
- [Uvi sales job posting] Uvi sales job posting | https://www.uvi.io/jobs/sales-agent/
- [Forbes, March 2026] The New Question On Deal Calls: "Who Owns This Model?" | https://www.forbes.com/sites/daraabasiita/2026/03/26/the-new-question-on-deal-calls-who-owns-this-model/
- [Uvi careers] Uvi careers page | https://www.uvi.io/jobs/head-of-customer-success/
- [Blocks & Files, October 2026] Uvi gives AI agents enterprise context | https://www.blocksandfiles.com/ai-ml/2026/10/14/uvi-gives-ai-agents-enterprise-context/5302514