The enterprise context gap that Atlan, Snowflake Horizon, and Google AI Mode have all exposed at once.
Something significant happened in May 2026. Google replaced its Search infrastructure — not updated it, replaced it. AI Mode crossed one billion monthly users. Queries are doubling every quarter. The signal was clear: the discovery layer of the internet now runs on context, not keywords.
At the same time, the enterprise data platform market reached its own inflection point. Snowflake announced Horizon Context. Atlan has been positioning for two years as the context layer for enterprise AI. Databricks has Unity Catalog. Microsoft has Purview. Every major data platform is now building the same thing: a governed, semantic layer that tells AI agents what internal data means.
The enterprise is investing heavily in making its data understandable — from the inside.
Nobody is governing the outside.
Two Context Layers. One Is Being Built. One Is Not.
When a data catalogue like Atlan enriches a table called crm.accounts, it adds descriptions, lineage, ownership, and glossary terms. An internal AI agent querying that table now knows what “account” means in your organisation — whether it refers to a parent company, a billing entity, or an individual contact.
That context lives inside the enterprise perimeter. It is consumed by Cortex Agents, Databricks Genie, or a BYO LLM connected via MCP. It is not visible to Google’s AI Mode. It is not available to ChatGPT when a prospect asks which vendors offer what your organisation sells. It is not accessible to the AI agent a potential partner is using to research your capabilities.
The external context layer — the one that governs what AI systems understand about your organisation, products, and services from the outside — is a different problem entirely. It is answered not by a data catalogue, but by a governed Schema.org knowledge graph: a structured, machine-readable representation of your public identity, published in JSON-LD, traversable by any AI agent, and maintained as a living document rather than a one-time SEO task.
Why the Timing Matters Now
Google I/O 2026 confirmed what VISEON has been building toward: AI discovery is the primary channel. The organisations that will be found, cited, and transacted with by AI agents are those whose external identity is explicitly declared in structured form — not those whose content is well-written, or whose paid media budget is large.
The enterprise data governance investment is not wasted. It is simply incomplete. An organisation that has spent two years building a governed internal context layer with Atlan has done the hard work of defining what its data means. Extending that investment to the external semantic layer — so that the same definitional rigour applies to the public-facing representation of the organisation — is the logical next step.
What the External Semantic Layer Covers
Where an internal data catalogue governs schemas, lineage, and metrics — the external semantic layer governs:
- Organisational identity — who you are, your legal entity, your relationships, your authoritative identifiers
- Products and services — what you offer, how it is typed, priced, and available
- Capabilities and credentials — certifications, partnerships, sector expertise
- Vocabulary — the defined terms your organisation uses, and what they mean to an AI agent encountering them for the first time
- Commerce readiness — BuyAction schema, MCP endpoints, ACP/UCP protocol readiness for agentic transactions
None of these are covered by a data catalogue. All of them are covered by a well-built Schema.org knowledge graph.
The Differentiator Is Structural, Not Cosmetic
The market’s immediate response to Google AI Mode has been to rewrite content — adjust copy, restructure pages, use language that AI systems respond to. This is the wrong answer to the right question. AI systems do not struggle to read marketing content. What they lack is the structured context that tells them what that content means and whether to trust it.
Content without context is inference. Context without content is still machine-readable. The enterprise already understands this for internal data. The same principle applies externally.
VISEON builds the external semantic layer — auditing, constructing, and maintaining the Schema.org knowledge graph that makes your organisation’s public identity explicit, governed, and AI-traversable. Each knowledge graph is underpinned by a semantic contract: a versioned, machine-readable definition of what your entities mean, what your products cover, and where the boundaries of your data lie — so AI agents encounter governed meaning, not inference. Without it, any LLM connecting to your API or MCP endpoint arrives with no context whatsoever: no understanding of your terminology, your product boundaries, or what a confident-sounding answer actually covers. It is the external catalogue your internal platforms assume exists — and the answer to the discovery problem that Google I/O 2026 made impossible to ignore.