The report names Kyvos among the semantic layer vendors shaping AI context

Wrong answers now top the list of generative AI concerns. In a Gartner Peer Community poll, 35% of respondents named inaccuracy as the threat weighing most on their minds, ahead of lack of security and bias at 21% each. Gartner analysts trace most AI application failures to insufficient or inaccurate context.

Gartner’s report, Data Intelligence Monthly: Executive Insights on Context for AI, draws on a study of vendors building context capabilities for AI and Kyvos is one of the vendors included in that study.

Gartner examines what enterprise AI needs from context and where organizations stand today. Its findings reflect a shift underway across the industry: semantic layers are expanding from BI to AI. This blog walks through main insights from the report, alongside Kyvos’ perspective on what they mean for enterprises building AI context.

Why does AI give conflicting answers to business questions?

When different systems define the same business term in different ways, AI answers diverge.

Take “top customer.” Sales defines it by lifetime revenue, support by ticket volume and finance by contract value. An agent or application that pulls from all three can return three defensible answers, none of them reconciled.

One common response is to connect more systems. Protocols such as MCP let agents access more tools and data, while larger context windows let models process more information. But neither resolves which definition of “top customer” the business stands behind.

Gartner’s executive Insight:

60%
Gartner projects that by 2028, 60% of agentic analytics projects relying solely on the Model Context Protocol will fail for lack of a consistent semantic layer beneath them.

What is AI context and what does it include?

Gartner’s executive insight:

Gartner describes AI context as the structured, dynamic and persistent information environment that lets AI models and agents interpret inputs, make decisions and act in line with business objectives. It identifies three core components of the context layer: semantics, operational state and provenance.

Semantics capture meaning. They define what data means and how it is measured, how customers link to accounts, how products roll into categories and which business rules apply. Ontology defines the concepts in a business domain and how they relate, while knowledge graphs and context graphs put that ontology into practice by linking real business entities.

Operational state supplies right-time awareness: current data about business entities, processes and conditions, along with organizational knowledge such as policies and documentation.

Provenance tracks lineage, decisions, actions and outcomes, so every answer can be traced and trusted.
Governance runs where these components meet, keeping definitions, ownership and access policies consistent across teams and AI agents.

Why are semantic layers central to AI context?

A semantic layer models, governs and serves business definitions to every tool or agent. That makes it the ideal foundation for building AI context.

Gartner’s executive Insight:

93%
Gartner projects that of D&A and AI leaders are either using or considering semantic modeling, such as taxonomies and ontologies, within the next 18 months to support AI and data science.

Once primarily associated with defining metrics for dashboards, semantic modeling now has a broader role: giving AI a consistent view of the business.

The shift makes practical sense. A semantic layer captures what the business has agreed on: how gross margin is calculated, which fiscal calendar applies and who can see regional sales figures. Those definitions have been tested against years of reporting and are enforced every time a dashboard runs a query. Building context for AI from scratch would mean rediscovering all of it.

AI raises the stakes on that work. A dashboard answers the questions its designers anticipated. An AI assistant or agent faces open-ended questions in everyday language and acts on the answers it generates. Each answer has to resolve to the same definitions a finance team would use, with the same access rules applied. A semantic layer grounds AI in those definitions, so the number in a board report matches the number an agent quotes.

Gartner’s executive insight:

Gartner predicts that by 2027, organizations that prioritize semantics in AI-ready data will increase agentic AI accuracy by up to 80% and reduce costs by up to 60%.

The payoff is significant and the direction is settled. The challenge is turning semantic modeling into context that works across enterprise data and AI applications.

How is the vendor landscape converging on AI context

Context for AI comes from several vendor technologies.

Semantic models hold the business definitions the others rely on.

Knowledge graphs model entities and their relationships.

Data catalogs contribute metadata and discovery.

Protocols such as MCP let agents reach external tools and data. Each adds something AI needs, and each stays consistent only when it draws on those definitions.

Semantic layers themselves are converging from different directions. Data warehouses and lakehouses now ship native semantic views. BI tools such as Power BI, Tableau and Looker carry their own semantic models.

Standalone semantic layer vendors sit between these two, serving every tool from one place. Their capabilities differ widely. Most cover metrics and dimensions, and those built into a warehouse or BI tool govern meaning only within that platform.

Complete business context needs relationships, entities and enterprise knowledge alongside metrics, a combination that most semantic layers were never designed to deliver.

Gartner’s executive insight:

44%
of CDAOs rank building a contextual semantic layer for agents and AI as a top-five priority, yet only 18% of data leaders are very confident they can deliver it.

That fragmentation explains the gap. Enterprises know what they need, but assembling it from scattered components remains hard.

How Kyvos delivers complete and connected AI context

Kyvos starts where most semantic layers stop. It extends governed definitions with the relationships, entities and enterprise knowledge AI needs for accurate, trusted answers.

Gartner’s executive insight:

Gartner names Kyvos among the semantic layer vendors building AI context platforms.

Semantic models define metrics and business logic once and enforce them at query time, so “revenue” resolves identically for every dashboard and agent. Policies, playbooks and product documentation are integrated alongside, so AI reasons with the enterprise knowledge that governs how the business runs.

Kyvos’ context graph connects entities through relationships, with causal paths and source bindings that let AI trace an answer back to its origin. Context is assembled dynamically for each request, with governance built into every step, so AI answers stay accurate and trusted.

In production, accurate answers also need to be fast and affordable at enterprise scale. Kyvos customers achieve:

100%AI accuracy
63%Token savings
1500xFaster queries
50%+Lower cloud compute

Semantic layers sit at the center of AI context, and Kyvos is one of the vendors bringing complete and connected context into production today.

Evaluating semantic layers for AI? Explore the Kyvos Semantic Layer Buyer’s Guide.

Frequently Asked Questions

What is business context for AI?
Business context is the information AI needs to understand how an organization operates. It includes business definitions, metrics, relationships, rules and enterprise knowledge that help AI interpret data in the right business context.
What does Gartner's research say about semantic layers for AI?
Gartner's Data Intelligence Monthly: Executive Insights on Context for AI highlights the growing adoption of semantic modeling for AI. It found that 93% of D&A and AI leaders are using or considering semantic modeling within 18 months to support AI and data science. It also found that 44% of CDAOs rank building a contextual semantic layer as a top-five priority, while only 18% of data leaders are very confident they can deliver one.
What is a context graph?
A context graph shows AI how business entities connect to each other. In Kyvos, it links KPIs, entities and enterprise knowledge through relationships and causal paths, with source bindings that let AI trace each answer to its origin.
What should organizations adopting enterprise AI focus on?
Organizations should focus on giving AI reliable business context. The priority is shifting from selecting more powerful models to grounding those models in governed definitions, which is what moves AI from pilots to production.
Does Kyvos work with existing BI tools and AI agents?
Yes, Kyvos serves the same governed context to BI tools, analytics applications and AI agents. A metric resolves identically wherever it is used, and the Kyvos MCP server lets agents reach that governed context through the protocol.