Jedify raises $24M to help companies arm AI agents with context on their business

AI vendors beforehand their endeavor products arsenic if they’re turnkey solutions, but the chances are debased that AI agents volition deed the crushed moving close away. Unless you enactment successful the effort to bid a exemplary connected the specifics of your business, it’s improbable to recognize however your company, for example, defines gross oregon knows who is allowed to spot which file. That’s portion of the crushed wherefore we’re seeing AI companies deploying engineers to assistance integrate their AI products into customers’ systems.

New York-based startup Jedify is attacking this precise gap. The institution says its level connects to enterprises’ cognition sources via APIs to physique a “context graph” astir their concern that AI agents tin usage to enactment better. These sources tin beryllium databases, information warehouses and lakes, SaaS apps oregon BI tools, arsenic good arsenic unstructured sources specified arsenic reports, documentation, codification bases, and adjacent Slack channels and gathering recordings.

To physique that out, Jedify has raised $24 cardinal successful a Series A backing circular led by Norwest, TechCrunch has exclusively learned. The circular saw information from returning backers S Capital VC and Cerca Partners, arsenic good arsenic caller capitalist Oceans Ventures. Data elephantine Snowflake besides participated arsenic a strategical capitalist and is integrating the startup’s tech with its AI products, specified arsenic its Cortex AI service, Semantic Views, and CoWork.

Jedify’s transportation is that to beryllium utile wrong enterprises, AI agents request entree to the relationships betwixt entities, data, permissions, domain knowledge, workflows, operational assumptions, and company-specific terminology. This context, the institution says, allows an AI cause to constrictive its attraction to the accusation that is applicable to a peculiar task alternatively of searching crossed everything a institution has.

Co-founder and CEO Assaf Henkin (pictured above, connected the acold right) pointed to Kiteworks, a compliance company, arsenic an illustration of however customers are utilizing Jedify. Kiteworks connected Snowflake, Tableau, Notion, and interior playbooks, including documents and screenshots, to Jedify, past built agentic tools for antithetic lawsuit workflows.

“They wanted to limb their sellers and relationship teams with a blase app — you tin deliberation of it arsenic some similar a dashboard exertion and a real-time conversational application. When they spell into a lawsuit conversation, Jedify builds for them, connected the fly, everything they request to know. And during the conversation, they can, successful existent time, get precise circumstantial details surfaced proactively,” Henkin said. 

Jedify’s discourse graph. Image Credits: JedifyImage Credits:Jedify /

Henkin argues that Jedify’s discourse graph is antithetic from the semantic layers, metadata catalogs, and cognition graphs that companies already usage due to the fact that it is multi-dimensional, capturing relationships crossed entities, data, people, permissions, and customers. It’s besides model-agnostic and updates successful existent clip arsenic accusation flows into and retired of the systems it is connected to.

“When you privation to alteration an agentic solution to truly beryllium autonomous, to thrust decisions crossed CRM data, Zendesk tickets, possibly telemetry information that’s coming successful existent time, that’s erstwhile a discourse graph is overmuch amended successful presumption of capabilities versus a semantic layer,” helium said.

Permissions are an evident hurdle here. It wouldn’t bash for an cause to springiness an intern entree to the CFO’s gross projections, for example. Henkin said his level works to code that by inheriting permissions from individuality systems, record systems, SaaS tools, and databases, including row-, column-, and table-level entree rules, past lets its customers make further groups that specify what and whom agents oregon workflows are allowed to reach. It besides offers observability and governance tools to assistance customers guarantee their AI agents are behaving arsenic intended.

Jedify is presently targeting mid-market and ample endeavor customers that person mature information stacks and aggregate databases oregon information warehouses. Henkin said the institution has betwixt 10 and 20 aboriginal customers, 1 of which is The Weather Company, and is seeing involvement from data-heavy sectors specified arsenic gaming, industrials, and user packaged goods.

Snowflake’s concern and concern are notable due to the fact that ample information platforms are besides trying to physique akin capabilities truthful their customers tin usage AI much efficaciously with their data. But Henkin argues that Jedify is complementary to specified efforts due to the fact that overmuch of a company’s information — and astir of its organization cognition — isn’t usually stored with a azygous unreality provider.

“[The ample information companies] volition archer you, ‘Oh yeah, conscionable bring everything.’ But successful reality, companies person aggregate databases, and warehouses, and information solutions […] The large happening is that not each of your information is successful those environments, and astir of your cognition is not there, truthful it’s a spot of a disadvantage that they really have,” helium said.

Henkin besides noted that for companies trying to bash this connected their own, grooming an AI exemplary to physique a comparable discourse furniture tin beryllium cost-prohibitive, particularly arsenic companies are scrutinizing and clamping down connected their AI token usage.

And the accelerated advances successful AI exemplary improvement play into the company’s broader bet: arsenic models turn much susceptible and much interchangeable, proprietary discourse that helps those models enactment amended wrong businesses volition beryllium a invaluable and durable moat.

The startup volition usage the caller currency for product development, hiring, and go-to-market motion. It brings the firm’s full backing to astir $33 million.