RAW PECAI MARKET & ECOSYSTEM ADVISORY

Battleground 03 · October 2026 edition

Who Decides What the Data Means?

Governing what data means to AI agents

Strategic Brief

September 11 – October 11, 2026 · 30-day retrospective · Evidence checked through October 11, 2026

The bigger trend

The contest over what data means is becoming a contest over who serves that meaning to every agent.

Data platforms are no longer only storing and governing data. They are now publishing curated business definitions and exposing them, under their own permissions, to agents built by other companies. That turns the context layer into a distribution point: whoever's definitions an agent reads first shapes its answer, even when the agent belongs to someone else.

Outside → In

What changed — Sense (verified)

Databricks, September 22–25. On September 22 Databricks introduced Unity Catalog Pages, a governed home for human-curated business definitions (owner, synonyms, description, related assets) that feed its Genie Ontology. Pages is in Beta. On September 25 Databricks made the Genie One MCP server generally available, so any MCP-compatible agent, including Claude, ChatGPT and Cursor, can ground data questions in Genie with Unity Catalog permissions enforced on every request. On September 24 sharing of governed metric views across accounts via OpenSharing also became generally available.

What it means — Interpret (RAW PEC view)

Databricks paired two moves: people define meaning once (Pages), and agents anywhere consume it under the platform's permissions (MCP server). The effect is to make the platform that holds the definitions the reference point other vendors' agents call. This sharpens, but does not settle, the contest with Snowflake's open-standard approach and Palantir's Ontology. A connector being available is not proof that one platform controls meaning, and a Beta feature is not evidence of adoption.

What to consider — Position

If your agents, or your customers' agents, will read business definitions from a platform you do not control, which definitions must you own, and can they move with you?

Inside → Out — suggested enterprise actions

Suggested actions for readers, not a report of actions taken.

Who needs to align — Align

Product, data and analytics, architecture, security and partnerships.

What the enterprise could execute — Execute

Choose one high-value term, such as "active customer." Define it once, then compare answers from a platform-native path and a portable path against the same set of questions.

How to measure and learn — Learn

Answer consistency across agents, permission correctness, time to integrate, and the effort required to move the definition.

Supporting competitive moves (this period)

  • Databricks · Build ·

    Introduced Unity Catalog Pages for human-curated business definitions in Genie Ontology (beta)

    Primary source · www.databricks.com

  • Databricks · Build ·

    Genie One MCP server generally available; outside agents can query Genie with Unity Catalog permissions enforced

    Primary source · docs.databricks.com

RAW PEC Perspective

The data contest is over who approves what a business term means to AI agents, because whoever holds meaning shapes what agents may do.

The management question

Who in your company approves what a word like "customer" means?

Why now

Recorded summaries

  • Snowflake / Databricks: context layers announced two weeks apart: Snowflake June 2 and Databricks June 16
  • Databricks: announced Genie Ontology context layer
  • Snowflake: announced Horizon Context

RAW PEC Perspective

The stored record includes product announcements and acquisitions involving data, semantics and context; these are different kinds of moves, not equivalent purchases of a context layer.

Attribution is as recorded and was not independently revalidated here.

Where this sits in the seven-layer stack

Contested layers: 3 Data, context and interoperability.

Adjacent layers: 5 Domain platforms.

  1. Layer 7Business outcomes and value realization
  2. Layer 6Workflow, agents and orchestration
  3. Layer 5Domain platformsAdjacent
  4. Layer 4Models and intelligence
  5. Layer 3Data, context and interoperabilityContested
  6. Layer 2Compute, cloud and security
  7. Layer 1Physical and energy foundations

Contested control points

  • Layer 3, Data, context and interoperability: Can the required context move if we change providers?

Contest dimensions

  • Provenance
  • Authority
  • Freshness
  • Meaning
  • Access

Decision lens

RAW PEC Perspective Data to agent action: enabling context and cross-cutting governance

Cross-cutting governanceMeaning, access, freshness and permitted actions across the whole process

Semantics and context — enabling layerDefinitions and relationships that support use across systems

Systems of recordBusiness data, including existing definitions and rules

Agents and actionsUse of data and context within governed permissions

This is a technical data-to-action lens, not the Corporate Loop. The recurring coordination cycle remains Sense → Interpret → Position → Align → Execute → Learn, with evidence from execution and learning feeding back to Sense. The canonical layer mapping above is unchanged.

Text equivalent: Systems of record can contain business meaning. Semantics and context provide an enabling layer for agent use across systems. Governance covers meaning, access, freshness and permitted actions throughout the data-to-action process; it is not a later step.

A qualitative reading aid. It is not a ranking, readiness score, measurement or statement about any participant.

Participants

These are overlapping roles, not exclusive categories or rankings. Notes are as recorded, not independently revalidated.

Established ontology platforms

  • Palantir (Incumbent): the Ontology; revenue up 93% in Q2 2026 Linked source
  • Microsoft (Fabric IQ) (Incumbent): an ontology inside its data platform, in preview

Data platforms moving up

  • Databricks (Challenger): Genie Ontology: context inferred from how the business works, plus human-curated definitions (Unity Catalog Pages, beta) served to outside agents over MCP
  • Snowflake (Challenger): Horizon Context and an open standard for shared definitions

Buying meaning

  • Salesforce (Informatica) (Incumbent): bought a data-management leader
  • ServiceNow (data.world) (Challenger): bought a data catalog and knowledge-graph company

Ecosystem moves

Each entry is shown as recorded, newest first. Type, date and status are kept separate. A reference appears only where one is recorded for that entry.

4 Build · 2 Buy · 1 Partner · 2 Compete

  1. Databricks

    Genie One MCP server generally available; outside agents can query Genie with Unity Catalog permissions enforced

    Move type
    Build
    Date
    Date meaning
    Date associated with the recorded entry (day precision); its event or publication basis is not specified here.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.

  2. Databricks

    introduced Unity Catalog Pages for human-curated business definitions in Genie Ontology (beta)

    Move type
    Build
    Date
    Date meaning
    Date associated with the recorded entry (day precision); its event or publication basis is not specified here.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.

  3. Snowflake / Databricks

    context layers announced two weeks apart: Snowflake June 2 and Databricks June 16

    Move type
    Compete
    Date
    Date meaning
    The stored June 16 date is the later announcement in the paired record. The summary separately retains Snowflake June 2 and Databricks June 16.
    Status
    Paired announcements as recorded; adoption and interoperability outcomes are not established here.
    Reference
    No entry-specific reference is recorded.

  4. Databricks

    announced Genie Ontology context layer

    Move type
    Build
    Date
    Date meaning
    Date associated with the recorded entry (day precision); its event or publication basis is not specified here.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.

  5. Snowflake

    announced Horizon Context

    Move type
    Build
    Date
    Date meaning
    Date associated with the recorded entry (day precision); its event or publication basis is not specified here.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.

  6. Snowflake / 54 vendors

    Open Semantic Interchange: a shared standard for business definitions

    Move type
    Partner
    Date
    Date meaning
    Date associated with the recorded entry (month precision); its event or publication basis is not specified here.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.

  7. Salesforce / Informatica

    Salesforce completes acquisition of Informatica

    Move type
    Buy
    Date
    Date meaning
    Date associated with the recorded entry (day precision); its event or publication basis is not specified here.
    Status
    Action milestone described as completed in the stored summary; operating outcomes and impact are not established here.

  8. ServiceNow / data.world

    ServiceNow signs definitive agreement to acquire data.world

    Move type
    Buy
    Date
    Date meaning
    Recorded month of signing the definitive agreement; not an inferred closing date.
    Status
    Definitive agreement signed as recorded; this entry does not establish that the acquisition closed.

  9. Databricks / Palantir

    overlapping approaches to enterprise context and ontology

    Move type
    Compete
    Date
    Undated
    Date meaning
    No date recorded for this entry; none is inferred.
    Status
    The recorded summary describes an action or announcement; completion and operating outcomes are not determined by this status label.
    Reference
    No entry-specific reference is recorded.

Evidence & RAW PEC Perspective

Recorded summaries and references

These summaries are not independently verified by this presentation. References are associated only with their mapped entries; a link does not establish support for every claim in a summary.

Snowflake / Databricks: context layers announced two weeks apart: Snowflake June 2 and Databricks June 16

No entry-specific reference is recorded; this item is not established here.

Databricks: announced Genie Ontology context layer

Recorded entry-specific reference

Snowflake: announced Horizon Context

Recorded entry-specific reference

Snowflake / 54 vendors: Open Semantic Interchange: a shared standard for business definitions

Recorded entry-specific reference

Salesforce / Informatica: Salesforce completes acquisition of Informatica

Recorded entry-specific reference

ServiceNow / data.world: ServiceNow signs definitive agreement to acquire data.world

Recorded entry-specific reference

Databricks / Palantir: overlapping approaches to enterprise context and ontology

No entry-specific reference is recorded; this item is not established here.

RAW PEC Perspective

RAW PEC Perspective: systems of record can already contain business meaning. Semantics and context can help agents use definitions and relationships across systems, while governance applies throughout the data-to-action process.

Management Implications: when vendors define context differently, clarify who in your company has authority over a term like customer and how those definitions are maintained.

Management interpretation, not an independently verified finding.

Evidence & Limitations

Participant notes and move summaries are shown as recorded and were not independently revalidated here.

Growth figures and announcements are recorded claims; adoption and outcomes are not established here.

Overlapping approaches are not evidence that one will prevail or that definitions are interoperable in practice.

Questions for your leadership team

Management Implications: questions to guide decisions, not verified findings.

  1. Who in our company approves what a word like customer means, and is that written down?
  2. Could our definitions and context move with us if we changed data platform?
  3. Which agent actions rely on meaning that nobody currently owns?

Related battlegrounds

Battlegrounds that share a contested or adjacent layer with this one.

Sources

  1. Business Wire — Palantir Q2 2026 results, August 3, 2026
  2. Snowflake — Build the interoperable lakehouse, June 2, 2026
  3. Atlan — Snowflake Summit 2026 announcements
  4. Atlan — Databricks Genie Ontology
  5. Databricks — Databricks Launches Genie One: All-New Agentic Coworker for Every Team, June 16, 2026
  6. Salesforce — Salesforce completes acquisition of Informatica, November 18, 2025
  7. Business Wire — ServiceNow Workflow Data Fabric ecosystem and data.world agreement, May 2025
  8. Databricks — Unity Catalog Pages: a governed home for your business knowledge in Genie Ontology, September 22, 2026
  9. Databricks — Platform release notes, September 2026

Evidence checked through:

Independent and unsponsored: no company pays to appear in these battlegrounds.

Discuss who owns data meaning in your company with Ron.

Talk with Ron