Ask a large company what its Q3 revenue was and you will get several answers, all of them defensible. Finance has one number, the sales dashboard has another, and somewhere in the warehouse there are four tables named some variant of `revenue_final`. Humans navigate this by knowing who to trust. AI agents do not know who to trust, which is why a Tel Aviv and New York startup called Euno just raised $23 million to build machine-readable institutional memory — the layer that decides which revenue number an agent is allowed to return.
The Series A was announced September 9 and led by N47, with participation from 10D, which led Euno's seed. The angel list is a tour of the Israeli enterprise-software diaspora: Wiz co-founder Yinon Kostika, Cyera co-founder and CEO Yotam Segev, Eon co-founder and CEO Ofir Ehrlich, Tavily founder Rotem Weiss, and Mark Nelson, the former CEO and president of Tableau. Total funding now stands at $29 million. The company, founded in 2023 by CEO Sarah Levy and CTO Eyal Firstenberg and incorporated as Delphi.io Inc., employs roughly 30 people across the US and Israel and says it intends to roughly double that by year-end. No valuation was disclosed. Neither was revenue, customer count, or any growth multiple — worth noting for a round pitched on enterprise traction.
What Euno actually sells
Euno reads metadata, not data. Its platform maps column-level lineage across warehouses, transformation tools and BI systems, then layers on usage patterns, ownership, business logic and governance state to build what the company calls a context graph. Assets get automatically labelled — "AI-ready," "PII-free," "certified" — according to rules the customer sets, and the labels update as the environment changes. A dashboard that was never certified but suddenly gets heavy traffic can trigger a workflow. PII propagating into a new table can trigger another.
The point of all that plumbing is the last mile: agents query the graph through a Model Context Protocol server, which means Claude, Cursor, VS Code Copilot and anything else speaking MCP can ask Euno what a metric means and whether the person asking is allowed to see it. The company has also built its own query language, Euno Query Language, and documented an integration with Snowflake Intelligence. Its connector list covers Snowflake, Databricks, dbt, Tableau, Looker, Power BI, ThoughtSpot, Sigma and Omni. Euno explicitly sits beside the existing stack rather than replacing it — a positioning choice that is both a sales advantage and a long-term vulnerability.
The pitch to buyers is time. Levy says Euno cuts the work of preparing a context layer for agentic deployment from as much as a year down to a few weeks, on the theory that most of what determines whether data can be trusted does not need to be hand-documented — it can be inferred from operational signals about how the data is actually used. That claim is unaudited, and "up to a year" is the kind of baseline that vendors get to choose for themselves.
Levy's framing of the strategic bet is more interesting than the product demo. "Models keep getting better and cheaper, but the long-term AI moat for enterprises will increasingly come from the accumulated record of how work gets done," she said. "Competitors may have access to the same frontier models, but they cannot easily replicate the proprietary context and experience an enterprise has accumulated through its own operations."
N47 partner Moshe Zilberstein put the investment case more bluntly: "For an AI agent to move the needle, it must act on current and trusted business data. That demands precise, scalable and contextual infrastructure. Demand for this is accelerating fast, and Euno is the only company building it AI-native, instead of retrofitting a solution meant for people."
Named customers are AlphaSense, the market-intelligence platform, and Zayo Group, the telecom infrastructure operator; Euno's own site documents a deployment at Bolt. The company was named a Cool Vendor in Gartner's 2026 Coolest Vendor Innovations in Data Management report, which credited the context graph with giving AI systems semantic understanding of data assets.
Why it matters
The semantic layer is an old idea with a new job. For two decades it was a BI convenience — define a metric once so Looker and Tableau agree. Then agents started issuing queries at machine volume, without the tacit knowledge an analyst carries, and the definition layer went from nice-to-have to the thing standing between an enterprise and an agent confidently reporting the wrong number to a board.
That makes it contested ground. dbt, Cube and AtScale sell vendor-neutral semantic layers. Snowflake ships semantic views and Databricks ships metric views through Unity Catalog, which means the two companies that already hold the data have every incentive to give the definition layer away. Looker's LookML and Power BI carry their own. Euno's answer is to operate a level above all of them — governing the estate rather than competing to define individual metrics — which works only as long as enterprises stay multi-vendor. Zilberstein's "only company building it AI-native" is investor language, not a market fact.
What to watch
Whether Euno discloses any revenue figure at the Series B, which will say more than any Gartner mention. Whether the headcount actually doubles, or whether the sales hiring outruns the deployments. Whether Snowflake and Databricks extend their native semantic layers into cross-platform governance — the move that would squeeze Euno hardest. And whether "context layer" survives as a product category at all, or gets absorbed into the catalog and warehouse vendors that already sell the surrounding pieces.
“Models keep getting better and cheaper, but the long-term AI moat for enterprises will increasingly come from the accumulated record of how work gets done.”— Sarah Levy, Co-founder and CEO, Euno