How Do You Get Claude To Talk To All Your Enterprise Data? >>> Read the blog by our CEO

February 12, 2026

The New Agentic Analytics Fabric OR How to Get Claude to “Talk To” All Your Enterprise Data

Three forces — the complexity of production-ready data agents, the unsolved challenge of context engineering, and the lack of open architecture choices — are converging to demand a fundamentally new approach to enterprise analytics.

 Prat Moghe

Prat Moghe

CEO

Three significant themes are around us in the enterprise data & analytics industry:

  1. After the internal OpenAI data agent was shared, there has been an explosion of interest in how to build agents to leverage enterprise data and context. Claude Plug-ins/Skills have further accelerated the space. The OpenAI article illustrates that building a production-ready stack that works for self-service even in a narrow modern environment is not easy. Think about a realistic Fortune 1000 enterprise that has 10X the scale and diversity across platforms, tools, and users. The complexity and scaling challenges are real. 
  2. Since Ashu Garg and Jaya Gupta proposed context graphs, we now have a vocabulary to describe the living and breathing process context behind outcomes. As an example, “Talking to business data” for insights is an unsolved problem today because we lack the institutional knowledge of repeatable context graphs that enable provable and scalable accuracy. No enterprise I have met has solved this problem at scale. 
  3. Meanwhile data teams are looking for the right post modern-data-stack architecture to support agent-driven analytics in the enterprise. Every enterprise is scoping home-grown agentic platforms or trying to expand platform-specific offerings. Quick solutions are not scalable in the long run. There are no good choices. The new architecture needs to be open, extensible and future proof in a space that changes weekly. 

That’s why I am excited about Kevin Petrie’s paper describing a new agentic analytics fabric architecture. The new stack scales the problem described by the OpenAI data agent article across any enterprise platform or tooling or business domain. Your favorite agent should be able to point to this stack (its MCP server) and build scalable and accurate insights with:  

  1. Federated and efficient live access to all your data regardless of the platform 
  2. Immediate and curated access to the relevant context across distributed contextual sources
  3. Personalized agent engineering with intent and memory across each domain
  4. Fine-grained access control that ensures governance from the user-level all the way to the data

Click here to read the whole paper.

Related Blog Posts

February 11, 2026

5 Ways to “Talk” to All Your Data With Promethium

Promethium Mantra™ lets analysts talk to all their enterprise data — across warehouses, SaaS apps, and on-prem systems — to get fast, trusted answers grounded in real business context.

Continue Reading »
January 20, 2026

The Context Engineering Challenge No One Talks About

AI accuracy doesn’t fail because models can’t write SQL — it fails because enterprises underestimate the cost and complexity of engineering business context at scale.

Continue Reading »
January 6, 2026

5 Predictions for Enterprise Data in 2026: When Agentic AI Goes to Production

2026 is when AI pilots have to become production systems — here are the five infrastructure shifts that separate the companies who scale from those stuck explaining why their AI investments haven't delivered ROI.

Continue Reading »