What makes enterprise data competitive if volume alone no longer matters?

Chris Meringolo, writing on Merkle's site after this year's Snowflake Summit, argues that data volume stopped being the enterprise's competitive advantage once agents began acting at machine speed; the scarce resource is now "meaning" — a durable, shared, machine-legible account of what the enterprise's own terms refer to. He frames the asymmetry with an example: "Ask a frontier model about the public world and it answers well. Ask why your private-label cold brew lost share in San Francisco last quarter, and the best it can do is guess," while "a regional manager could reconstruct that answer in a week" because she holds "a single, coherent model of how the pieces fit together" that no one system captures — she is, in his words, "the ontology your enterprise lacks."

The mechanism he gives is that enterprise meaning is distributed and contested rather than simply missing: "Sales' active customer is not Finance's. There is no single resolver." A human papers over this with a clarifying question; an agent cannot, because "an autonomous agent never gets the clarifying turn a human gets," so definitions have to be resolved in advance or agents act on the wrong one. His proposed sequence runs data (governed, in open formats like Iceberg) to semantics (shared metric definitions, citing the Open Semantic Interchange spec finalized with Snowflake, Salesforce, and dbt Labs in January 2026) to ontology (relationships and possible actions encoded once and exposed through a knowledge graph). He cites Snowflake's own "Ontology on Snowflake" materials, which report that "ontology-grounded agents have scored 10 to 20 points higher in accuracy than a semantic-view baseline."

This sharpens Does agent capability matter more than coordination infrastructure?: where that note names coordination, accountability, and evidence as the binding constraint on agentic systems generally, Meringolo names a specific artifact — an explicit, shared ontology — that a single enterprise must build before its agents can act coherently on internal business questions at all. It also gives a concrete instance of why Why do AI agents fail at workplace social interaction? finds private knowledge domains among the hardest failure modes for agents: the knowledge is hard not because it is absent but because it is "scattered across systems never built to talk to each other" and across thousands of employees who do not agree on what it means.

The piece is argument and vendor framing rather than independent measurement. The 10-to-20-point accuracy figure comes from Snowflake's own materials rather than a third-party study, the Open Semantic Interchange spec is new enough (finalized January 2026) that its adoption is unverified, and Meringolo writes for Merkle, a data and marketing consultancy with a commercial stake in enterprises commissioning exactly this kind of ontology-building work. What the excerpt establishes is a plausible mechanism and a named shift in emphasis from Snowflake Summit discourse — that agentic scale makes implicit human context a liability — not an independent demonstration that ontology investment pays off.