Jedify and Snowflake: $24 Million for Contextual AI

AI suppliers often tout their solutions as ready-to-use for businesses, but the reality is often more complex. For these AI agents to be truly operational, it is crucial to train them on the specifics of each company. Without this, they may fail to understand fundamental concepts such as the definition of revenue or file access permissions. This is why many AI companies deploy engineers to integrate their products into client systems.

Jedify, a New York-based startup, is tackling this issue by offering a platform that connects to companies' knowledge sources via APIs. This platform builds a "context graph" of the company's activities that AI agents can use to enhance their performance. Knowledge sources include databases, data warehouses and lakes, SaaS applications, BI tools, as well as unstructured sources like reports, documentation, codebases, Slack channels, and meeting recordings.

To develop this technology, Jedify raised $24 million in a Series A funding round led by Norwest. Among the investors are S Capital VC, Cerca Partners, and a new investor, Oceans Ventures. Snowflake, a data giant, also participated as a strategic investor and is integrating Jedify's technology with its AI products, such as its Cortex AI service, Semantic Views, and CoWork.

Jedify asserts that for AI agents to be effective, they must have access to rich context that includes relationships between entities, data, permissions, domain knowledge, workflows, operational assumptions, and company-specific terminology. This context allows AI agents to focus on relevant information for a given task, rather than sifting through the entire company's data.

Assaf Henkin, co-founder and CEO of Jedify, cited Kiteworks, a compliance company, as an example to illustrate the use of their technology. Kiteworks connected tools such as Snowflake, Tableau, Notion, and internal manuals to Jedify, enabling the creation of agent tools for various customer workflows. "They wanted to equip their sellers and account teams with a sophisticated application — think of it as both a dashboard application and a real-time conversational app. When they enter a conversation with a client, Jedify builds for them, on the fly, everything they need to know. And during the conversation, they can, in real-time, get very specific details proactively highlighted," Henkin said.

Henkin claims that Jedify's context graph is unique compared to existing semantic layers, metadata catalogs, and knowledge graphs because it is multi-dimensional. It captures the relationships between entities, data, people, permissions, and clients, and is model-agnostic, updating in real-time as information flows through connected systems.

Managing permissions is an obvious challenge. It would be unacceptable for an AI agent to grant an intern access to the CFO's revenue projections, for example. Henkin explained that Jedify's platform addresses this issue by inheriting permissions from identity systems, file systems, SaaS tools, and databases, including row, column, and table-level access rules. Clients can also create additional groups to define what agents or workflows can access. Jedify also offers observability and governance tools to help clients ensure their AI agents behave as expected.

Jedify is currently targeting mid-sized and large companies that have mature data stacks and multiple databases or data warehouses. The company has between 10 and 20 early clients, including The Weather Company, and is seeing interest in data-rich sectors such as gaming, industry, and packaged consumer goods.

Snowflake's investment and partnership are notable as large data platforms are also trying to develop similar capabilities. But Henkin argues that Jedify is complementary to such efforts, as a significant portion of a company's data, and most of its institutional knowledge, is typically not stored with a single cloud provider. "[Large data companies] will tell you: 'Oh yes, bring everything.' But in reality, companies have multiple databases, warehouses, and data solutions [...] The main issue is that all your data is not in those environments, and most of your knowledge isn't there either, so it's a bit of a disadvantage they actually have," he said.

Henkin also noted that for companies trying to do this themselves, training an AI model to build a comparable context layer can be prohibitively expensive, especially as companies review and restrict their use of AI tokens. Rapid advancements in AI model development also play into the broader bet of the company: as models become more capable and interchangeable, a proprietary context that helps these models perform better within companies could prove to be a valuable and sustainable asset.

The startup will use the recent funds for product development, recruitment, and go-to-market efforts. This brings the company's total funding to approximately $33 million.