Walmart’s SuperAgents: Why Semantics and Knowledge Graphs Are the Real Foundation

A story that caught my eye the other week was Walmart’s new agentic framework, their so-called SuperAgents. The move is a direct response to a common challenge in many organisations: the rapid - and often uncontrolled - proliferation of individual AI tools across the business.

What stood out was their clear advocacy for higher-level orchestrating agents. The framework consolidates dozens of tools into four primary agents:

🔹 Sparky: for customers, handling product recommendations and review summaries
🔹Associate: for employees, automating HR and Sales tasks
🔹Marty: for suppliers and partners, streamlining catalogue management and analytics
🔹Developer: for internal tech teams, enabling rapid creation of smaller “nano agents”

These aren’t standalone bots; they’re orchestration layers.

From a Knowledge Graph perspective, this is compelling for two key reasons.

First, if you're going to build a genuinely agentic framework, any two agents that need to coordinate must share a semantic understanding. As they interact, it’s essential that they use language in a way that’s consistent and contextually accurate. This is where a shared internal ontology, encoded in a knowledge graph, becomes invaluable: it grounds agents in a single source of enterprise truth, preventing errors and ensuring terms carry precise, agreed-upon meaning across the system.

Second, Walmart’s investment in semantics runs deep. Years ago, they shared work on their Retail Graph - a product knowledge graph built to power semantic search and product discovery. They even presented case studies on using graph databases for real-time recommendations as far back as 2013. More recently, they hired Daniel Danker as EVP of AI. He joined from Instacart - another company that has leaned heavily on a knowledge graph to tackle challenges like data quality scoring and out-of-stock replacement suggestions.

While I have no inside knowledge of Walmart’s current architecture, these signals point to something I’ve written about before: the AI iceberg.

No organisation wants a sprawl of disconnected AI tools. Yet when planning their own AI strategies, it’s easy to get distracted by the glittering surface - the magic of ‘super agents’.

Beneath that surface lies the real foundation: structured data, clearly defined semantics, and - in Walmart’s case - a long-term investment in knowledge graphs they now seem to be doubling down on.

That foundational work - aligning semantics so agents can collaborate, linking data so facts can be reliably retrieved - is the hard part. It’s also where the real value lies. Once that’s in place, layering AI on top becomes, relatively speaking, the 'easy bit'.

That’s the deeper lesson hiding just beneath what shows on the surface.

⭕ SuperAgents: https://www.linkedin.com/pulse/all-agents-suresh-kumar-lhxfc/

⭕ AI Iceberg: https://www.knowledge-graph-guys.com/blog/the-ai-iceberg