Gartner D&A Summit 2026: Key Takeaways on Context & AI

The context layer mandate: treating context as critical infrastructure

Adam Ronthal, VP Analyst at Gartner, and Georgia O’Callaghan, Director Analyst at Gartner, declared context the “new critical infrastructure” in their opening keynote. Organizations face a fundamental challenge: AI agents cannot operate reliably without understanding business context beyond raw data.

Four out of five organizations increased AI investments in 2026, yet only one in five shows measurable ROI. The gap stems from fragmented context spread across documentation, tribal knowledge, and disconnected tools. Without clear context, large language models make incorrect assumptions and decisions.

Gartner’s context engineering framework positions context as “the brain for AI” delivering three benefits:

Improved accuracy — Agents access the right information at decision time rather than reasoning without guardrails
Lower costs — Models reduce unnecessary exploration by operating within defined boundaries
Operational trust — Business rules and constraints enforce automatically through governance

Semantic layers form the foundation of the broader context layer needed for AI agents to function reliably. Modern platforms address this mandate by unifying semantic and context layers into a single enterprise context layer. Rather than maintaining separate systems for cataloging, definitions, and governance, organizations can operate from a unified context graph that AI agents query in real-time through standard interfaces like the Model Context Protocol.

AI-ready data: the 60% failure rate organizations must address

Roxane Edjlali, Senior Director Analyst at Gartner, explained AI-ready data is always contextual and use-case dependent. Organizations cannot rely on one-time data quality checks. Teams must continuously verify models still run contextually in the same conditions where they were developed.

Gartner identified three pillars organizations must adopt:

AI ambition — Establish clear goals for AI initiatives to generate a return on intelligence
Strong AI foundations — Build trusted data systems and governance to achieve a return on integrity
People empowerment — Enable employees with skills and tools needed for AI transformation

More than half of IT leaders worry about cost overruns related to AI initiatives. However, fewer than one out of five data and analytics or AI leaders believe uncertain costs will limit AI value. Only 44% of organizations have adopted financial guardrails or AI FinOps practices.

The Gartner AI-Ready Data Framework establishes three critical dimensions:

Alignment — Ensures data accessibility, semantics, accuracy, and lineage
Qualification — Validates continuous data quality for AI models
Governance — Manages AI lifecycle practices including responsible AI principles and policy enforcement

Organizations using specialized governance tools will decrease compliance costs by 20% by 2028 through automation and runtime enforcement.

Data governance and AI governance: finding common ground

Andrew White, VP Analyst at Gartner, and Lauren Kornutick, Director Analyst at Gartner, explored whether AI governance should extend from data governance or exist as a separate discipline. The audience split roughly 50/50, reflecting real organizational uncertainty.

Both analysts agreed on one critical point: if data governance fails, AI governance is at great risk. Gartner predicts 60% of organizations will fail to realize AI value due to poor integration of data and AI governance programs.

Organizations should connect data and AI governance through three overlapping areas:

AI-ready data and output quality — Both govern data inputs to AI systems and AI-generated outputs
Regulatory compliance — Shared responsibility for data regulations and emerging AI regulations
Culture change and adoption — Joint effort required for organizational transformation and user adoption

Data nutrition labels gained traction as a mechanism for documenting AI readiness. These labels provide context about data provenance, quality, and fitness for specific AI use cases in machine-readable formats that AI systems can consume.

Modern governance platforms serve both domains by treating context as shared infrastructure. AI governance and data governance operate from the same metadata foundation, ensuring policies apply consistently whether governing data assets or AI agents accessing those assets through the context layer.

Semantic layers and knowledge graphs: the foundation for reliable agents

Gartner positioned semantic layers and knowledge graphs as foundational infrastructure for agentic AI. Andrés García-Rodeja’s session introduced critical distinctions between knowledge graphs and context graphs:

Knowledge graphs:

Store entities and relationships using declared schemas and ontologies
Focus on “what” and “who” questions with relatively static content
Updated periodically with domain knowledge

Context graphs:

Capture decision traces and directional workflows from execution
Track “how” and “why” questions with continuously evolving content
Updated in real-time through actual system behavior

Organizations need both layers working together. Knowledge graphs provide the semantic foundation defining business entities and metrics. Context graphs layer operational state, behavior patterns, and decision lineage. Together they form the comprehensive organizational memory AI agents require.

The Model Context Protocol received extensive discussion as the emerging standard for connecting AI agents to data systems. However, García-Rodeja warned 60% of agentic analytics projects relying solely on MCP will fail by 2028 due to lack of a consistent semantic layer.

Data and analytics leaders can avoid the “black box” problem through three approaches:

Prioritize explainability for compliance
Build trust with transparent insights
Ensure traceability with semantic layers

Context engineering in practice: Gartner’s implementation framework

Gartner provided practical guidance for data and analytics leaders to implement context engineering immediately. Organizations should assess data readiness for agentic AI by reviewing existing semantic technologies and evaluating team skills in knowledge engineering.

Within 90 days, teams should execute the following:

Select pilot use case and define success metrics — Choose one high-value domain with clear business impact
Assemble multidisciplinary teams — Combine data engineers, domain experts, and governance leaders
Select technology components — Choose semantic layers and graph databases that work together as unified context layer infrastructure
Build semantic layer, ontology, and knowledge graph for the pilot domain — Start with one well-defined business domain

Within 12 months, organizations should achieve these milestones:

Evaluate pilot outcomes against success metrics
Deploy to production with proper monitoring
Expand the context layer to new domains
Establish ongoing governance measuring impact

The implementation requires sophisticated technical infrastructure including robust connectors for observability, advanced semantic understanding capabilities, and persistent enterprise memory layers that include process context.

What Atlan observed on the ground at Gartner D&A Summit 2026

“Context layer” was inescapable at this year’s summit. It came up in keynotes, in hallway conversations, and unprompted at booths across the expo floor. For some attendees it was a phrase they’d heard for the first time that morning. For others it was already a top priority with a budget attached. But regardless of where anyone fell on that spectrum, three patterns kept surfacing that reveal where most organizations actually stand.

1. The fragmentation problem is worse than leaders realize

Most data and analytics leaders we spoke with believed they had “decent metadata management” in place. Dig a little deeper, and a different picture emerged. Their context lives in 8–12 disconnected tools:

Business definitions in Confluence or SharePoint
Technical metadata in their data catalog
Lineage in separate tools (when it exists at all)
Data quality rules in observability platforms
Semantic definitions in BI tools or dbt
Governance policies in GRC systems

This fragmentation isn’t just inconvenient. It’s the reason AI agents fail. When an AI analyst needs to understand “revenue,” it can’t reconcile three conflicting definitions across systems. The AI hallucinates because it has no single source of truth for context.

What made this pattern striking was how consistent it was across very different roles. Data engineers building conversational analytics, governance leads focused on policy compliance, data leaders rethinking how semantic layers fit into their AI strategy — all of them described versions of the same bottleneck. The tooling heterogeneity across most enterprises has created a ceiling that no amount of model fine-tuning can break through.

2. Every new AI use case triggers the same painful rebuild

The second pattern was exhaustion. Data teams are burned out from context rework. Every new AI initiative triggers the same cycle:

AI team requests access to customer data
Data team scrambles to document what fields mean
Manual context creation takes 2–3 weeks
By the time context is ready, requirements have changed
Repeat for the next AI use case

In conversation after conversation, this showed up as three distinct failure modes. The Cold Start problem — teams that can’t get their first AI use case off the ground because no context exists. Testing Hell — a working prototype that can’t reach production because context is too brittle. And the Scaling Wall — teams that got one AI agent working but can’t replicate it without rebuilding context from scratch.

This isn’t a people problem. It’s an infrastructure problem. Without a unified context layer, organizations are forced into one-off context engineering for every agent they deploy. And many of the data leaders we spoke with are feeling it in a deeper way — their role is shifting from owning data to owning governed context for AI. That reframe resonated more than almost anything else we discussed.

3. The winners are treating context as infrastructure, not documentation

Organizations succeeding with agentic AI made a fundamental shift. They stopped treating context as documentation you create and started treating it as infrastructure you operate.

The difference shows up in three ways:

Context is automated, not manual — Successful organizations use AI to bootstrap context from existing patterns and documentation. They generate ontologies from their enterprise data graph rather than hand-coding them. Teams refine rather than create, cutting time from weeks to days.
Context is unified, not fragmented — They operate from a single context graph that AI agents can query. One source of truth that eliminates conflicting definitions. And they’re investing in connector depth, pulling context from every tool in their estate, not just the obvious ones.
Context is live, not stale — They serve context through standard interfaces like Model Context Protocol, keeping it open and portable so context becomes organizational IP — not locked inside a single vendor. It updates continuously as systems change.

What separated these organizations from everyone else? They think about context as a pipeline, not a project. From ingestion and enrichment, through engineering and curation, to portability across every AI agent in the enterprise. That pipeline or infrastructure mindset is what turns “AI-ready data” from a conference talking point into a real operational capability.

Learn more about Enterprise Context Layer

What this means for your 2026 roadmap

If your organization is still treating context as documentation rather than infrastructure, Gartner’s summit made clear: you’re building on a foundation that won’t scale to agentic AI.

The question isn’t whether to build a context layer. The question is whether you’ll build it as fragmented documentation or unified infrastructure.

Explore how your organization can build the context layer needed for AI agents that deliver reliable, governed business value.

Book a demo

FAQs about Gartner Data & Analytics Summit 2026

What is a context layer and why did Gartner emphasize it?

A context layer is infrastructure between enterprise data and AI systems that encodes business meaning, relationships, rules, and decision patterns. Gartner emphasized context because AI agents cannot operate reliably without understanding business context beyond raw data. Organizations currently have fragmented context across documentation and tribal knowledge, which causes AI systems to make incorrect assumptions.

What percentage of AI projects will fail without proper data readiness?

Gartner predicts 60% of AI projects will be abandoned through 2026 if organizations don’t establish proper AI-ready data practices. Only 37% of organizations currently have confidence in their data management practices for AI. Organizations must move beyond static quality checks to continuous contextual validation of data fitness for specific AI use cases.

How do semantic layers and context graphs differ?

Semantic layers define what metrics mean and how to calculate them consistently. They answer “what” and “who” questions using relatively static, declared schemas. Context graphs capture how decisions are made and why processes exist through continuously evolving decision traces. They answer “how” and “why” questions. Organizations need both working together for AI agents to act appropriately.

What is the Model Context Protocol and why does it matter for agentic AI?

Model Context Protocol (MCP) is an open standard for connecting AI applications to external data sources and tools, introduced by Anthropic in November 2024. MCP provides a standardized way for AI agents to access enterprise context rather than requiring custom integrations. However, Gartner warned that MCP alone is insufficient. Sixty percent of agentic analytics projects relying solely on MCP will fail by 2028 due to lack of a consistent semantic layer.

How much will organizations spend on AI governance platforms by 2030?

Gartner projects AI governance platform spending will reach $492 million in 2026 and surpass $1 billion by 2030. This growth is driven by increasing regulatory requirements and the need for specialized tools to enforce AI policies at runtime. Organizations using effective governance technologies can reduce regulatory compliance expenses by 20% through automation.

Should AI governance be separate from data governance or integrated?

The summit debate revealed organizations are split on this question. The practical answer is both governance domains need integration points around context, trust models, and shared policies while maintaining separate scopes for specific responsibilities. Data governance focuses on making data fit for purpose. AI governance addresses AI-specific risks including vendor management, model lifecycle, and ethical use. Both require context to function effectively.

Share this article

Gartner Data & Analytics Summit 2026: Key Takeaways

Key takeaways

What were the key themes and takeaways from Gartner Data & Analytics Summit 2026?

Key insights from the summit: