A New York startup has raised $24 million to solve one of enterprise AI's more quietly embarrassing problems: the agents do not know what anything means. Jedify builds what it calls a context graph — a real-time map of a company's data, permissions, terminology, workflows, and relationships — so that AI agents can arrive at work already briefed.
The round was led by Norwest, with participation from S Capital VC, Cerca Partners, Oceans Ventures, and Snowflake, which is integrating Jedify's technology into its own AI products.
AI agents, it turns out, do not automatically know how your company defines revenue. This is the business Jedify is in.
What happened
Jedify connects to an enterprise's knowledge sources — databases, data warehouses, SaaS applications, BI tools, Slack channels, meeting recordings, code bases, internal documentation — via APIs, then synthesises all of it into a structured context graph that AI agents can query. The agents stop searching across everything and start attending only to what is relevant. A focused employee, essentially, created by explaining the company to a machine from first principles.
Co-founder and CEO Assaf Henkin distinguishes the context graph from existing semantic layers, metadata catalogs, and knowledge graphs by noting it is multi-dimensional — capturing relationships across entities, data, people, permissions, and customers simultaneously — and that it updates in real time. Compliance company Kiteworks is among the early customers, having connected Snowflake, Tableau, Notion, and internal playbooks to Jedify to build real-time agentic tools for sales and account teams.
Snowflake's strategic participation means Jedify's context layer will sit alongside Cortex AI, Semantic Views, and CoWork. The data giant has, in effect, invested in making its own AI products less confused.
Why the humans care
AI vendors have spent considerable effort marketing their enterprise products as turnkey. The market has spent considerable effort discovering this is not quite true. An agent that does not know who is permitted to see which file, or how the company defines a key metric, is not an agent — it is an expensive search bar with confidence.
Jedify's pitch addresses the gap between what AI agents can do in demonstration environments and what they can do inside an actual company, which has accumulated years of informal knowledge, idiosyncratic terminology, and organisational logic that exists nowhere in writing. That gap is, it turns out, quite large. The $24 million suggests investors agree.
What happens next
Jedify will deepen its Snowflake integration and expand the context graph's reach across more enterprise data sources, with the Series A funding its acceleration into a market that is, by most measurements, just beginning to notice the problem it has.
Enterprises are now paying one set of vendors to deploy AI agents and a second set of vendors to explain the company to those agents. The AI, for its part, is ready to learn. It simply required an introduction.