If you’ve been anywhere near enterprise AI in the last six months, you’ve heard about context graphs. The concept, popularized by Jaya Gupta and Ashu Garg at Foundation Capital, gave a name to solving a challenge data practitioners have been wrestling with for years: enterprise application failing because the platforms don’t know anything about the business they’re supposed to serve.
For an agentic workflow to work at enterprise scale, it needs a deep understanding of the business. Which definition of revenue should apply? Which customer record is current? Which operational assumptions actually matter? Unless that context is available at runtime, the models will fail over and over again.
For an agentic workflow to work at enterprise scale, it needs a deep understanding of the business. Which definition of revenue should apply? Which customer record is current? Which operational assumptions actually matter?
This problem has existed for decades, yet the proliferation AI agents have made it impossible to ignore. The harder question inevitably became who would actually build the infrastructure to solve it. En masse, a wave of companies attached the context graph label to existing products – most without meaningfully addressing the underlying problem.
Today, we are excited to announce Norwest is leading the $24M Series A investment in Jedify, the company addressing this problem head on.
Why the Context Problem Has Proven So Hard to Solve
Enterprise data is fragmented across warehouses, CRMs, financial systems, BI tools, Slack, meeting recordings, and dozens of other sources. Each system carries its own definitions, schema, and assumptions about what a metric means. AI agents operating across this landscape either hallucinate because they lack the right context, or waste tokens processing irrelevant information.
The traditional response has been to build a semantic layer, which is a structured model mapping business meaning onto underlying data. But semantic layers built the old way require armies of data engineers and decay as the underlying data changes. As a result, they have to be rebuilt from scratch for every new agent or workflow. Seventy percent of the enterprises Jedify has spoken with have already tried and abandoned this approach.
What was missing was an autonomous semantic layer that could build and continuously maintain itself.
Jedify’s Semantic Fusion: A Context Graph Built for Enterprise AI Agents
Jedify’s core technology, Semantic Fusion, autonomously builds a customer-specific context graph on top of an enterprise’s existing data infrastructure. It ingests structured data from warehouses, CRMs, and financial systems alongside unstructured knowledge from documents, Slack, meetings, and BI dashboards, fusing them into a continuously updated, AI-ready model that deeply understands how the business actually works.
Rather than requiring data teams to author entities table-by-table, Jedify mines query logs at scale to infer how an organization actually uses its data. It parses BI dashboards to extract metric definitions, detects inconsistencies across sources, and uses existing dashboards as ground truth to auto-tune accuracy.
Rather than requiring data teams to author entities table-by-table, Jedify mines query logs at scale to infer how an organization actually uses its data.
With every interaction, the model becomes smarter. The context graph is a compounding, proprietary asset: the longer it runs, the more accurate it gets, and the harder it is to replicate. Customers choose their own tradeoff between speed and precision, and the system adapts. Jedify is model-agnostic by design, so enterprises are not locked into any single vendor’s infrastructure, including data warehouse providers who have a natural incentive toward token-intensive, platform-dependent solutions.
One way this is already making a difference is through Jedify’s deep integration with Snowflake’s Semantic Views, Cortex Analyst, and Snowflake Intelligence, demonstrating the multiplied impact when the context layer lives where enterprise data already exists.
What Customers Told Us About Jedify
We spoke with data leaders across financial services, media, and fintech during diligence, and heard consistent signals about the strength of Jedify’s platform. These leaders knew they needed a semantic foundation and were evaluating solutions that could build it without adding headcount or indefinite engineering cycles to maintain it. The autonomy of construction that Jedify offered was the deciding factor in nearly every conversation.
The Weather Company started using Jedify to scale insights and responses reliably. They went from 40-45% accuracy out of the box to over 85% after a focused refinement process with low-touch human intervention. Their assessment was that the switching cost grows not because of deep technical integration, but because of the investment in building the model, a compounding moat.
Another customer, a fintech company, ran a rigorous proof of concept with accuracy requirements at the 98-99% level and came away convinced. Their deployment has since expanded across product management, finance, and leadership, with AI agents for proactive metric monitoring on the roadmap.
The third-party technical validation we received reinforced our thesis that the key shift Jedify represents is starting from business context rather than schema.
Built to Compound: Why We Invested
Jedify is purpose-built for data-intensive agentic applications and workflows, the environments where context is most complex, most fragmented, and most consequential to get right. The core buyer is any enterprise team trying to move AI from prototype to production and needs a rich context, semantic layer.
CEO and co-founder Assaf Henkin previously built Kontera Technologies, a large-scale intelligence platform processing social graph and mobile data, acquired by Singtel in 2014. Building autonomous data pipelines at scale is not new territory for him or his co-founders Adi Elimelech and Eric Shani. It is the foundation the product was designed on, and it shows up directly in how Jedify constructs and maintains its semantic models without requiring a dedicated team to do it by hand.
The model providers who are supposed to solve this problem have a fundamental conflict of interest. They benefit from the most token-intensive, platform-dependent solutions. Jedify is independent and model-agnostic, and its incentives are fully aligned with the customer. The more accurate and efficient the context graph, the more value it delivers.
Every interaction, every new data source, and every workflow built on top of Jedify makes the asset more valuable and harder to replicate. That is the kind of infrastructure layer we look for, and we are proud to back the team building it.
To learn more about Jedify and see how the platform works, visit jedify.com. If you’re interested in joining the company, the team is hiring across the globe.
If you’re building in enterprise AI infrastructure, reach out to Assaf at [email protected] or Nikhil at [email protected].
