I just got back from Gartner Data & Analytics Summit 2026 in Orlando. 

And this year felt different. For the last two years, the conversation was all about GenAI. Bigger models. Faster outputs. More pilots.

This year, the tone shifted. The industry is starting to admit what’s actually broken. AI is not failing because of the models. It’s failing because enterprises don’t have context.

I also interviewed Atlan Team to understand this better – 

Interview with Prukalpa ⚡, Co-Founder & Co-CEO, Atlan

Interview with Austin Kronz, Head of AI & Data Strategy, Atlan

The real bottleneck: context

Across keynotes, analyst sessions, and side conversations, one theme kept repeating:

AI systems don’t understand your business. Not because they’re not powerful enough.

But because they lack context.

• What does “customer” mean in your company?

• Which revenue metric is the real one?

• Which data can an AI safely use?

Without clear answers, AI doesn’t reason. It guesses.

And that’s the gap most enterprises are dealing with right now.

1. Context is now critical infrastructure

The opening keynote made it explicit.

Context is becoming the brain behind AI.

• 4 out of 5 organizations are increasing AI investment

• Only 1 out of 5 are seeing real ROI

That gap comes down to fragmented context.

It lives in:

• Docs

• Dashboards

• People’s heads

• Disconnected tools

Gartner also shared a strong prediction:

60% of agentic analytics projects that rely only on MCP, without semantic foundations, will fail by 2028.

This is a big shift.

We are moving from building models → to building meaning.

2. Context is now a board-level problem

One of the most important sessions focused on where this is going long term.

By 2030:

• Almost all IT work will involve AI

• AI-first companies will outinvest others by up to 4x in data and governance

• The gap between leaders and laggards will become structural

Here’s the key insight:

If your AI is making decisions without governed context, you are introducing risk into your business.

Not small risk. Institutional risk. This is no longer a data team problem. This is a leadership problem.

3. Agentic AI depends on shared context

Another big signal: context got its own dedicated track. And the message was consistent across sessions: You cannot scale AI agents without a shared context layer. What leading teams are doing:

• Combining knowledge graphs, ontologies, and semantic layers

• Building shared context instead of one-off pipelines

• Using metadata as a control system, not just documentation

One example stood out:

A large enterprise used GenAI to enrich hundreds of thousands of data attributes and reduced manual effort by 60%.

But the real takeaway wasn’t the tooling.

It was this: AI scales only as fast as your context does.

4. Atlan was at the center of this shift

This is where things got interesting. Atlan’s positioning around a “context layer for AI” wasn’t just relevant. It matched exactly where the industry landed.

And it showed. Atlan ranked #1 in the Innovation Solution Showcase.

That’s not analysts. That’s practitioners voting.

At the same time:

• AI governance platform spend is expected to reach $492M this year

• And grow to $1B by 2030

Organizations are already assigning ownership of this layer to:

• Chief AI Officers

• CTOs

• CROs

This is no longer experimental. It’s becoming a core part of the stack.

What this means for you

If you’re investing in AI right now, here’s the shift to understand: The advantage is no longer in the model. It’s in the context behind the model.

Before you spend more on:

• New models

• More pilots

• More tooling

Ask:

Do we have a shared, governed understanding of our data? Because that’s where both value and risk now live.

How Atlan is approaching this

Atlan is building a unified context layer for both humans and AI agents.

The idea is simple: Create one place where business meaning, data context, and AI systems come together.

With:

• A control plane (Context Studio) to create and manage context

• An open metadata foundation

• Interoperability so you are not locked into one ecosystem

This is not another data tool.

It’s a new layer in the stack.

Check out more about how they’re building the context layer for AI

Go deeper

If you want to understand this space better, here are a few good starting points:

• Enterprise Context Layer Hub

• 2026 Guide to Implementing a Context Layer

• Context Studio product walkthrough

Gartner D&A 2026 was a reality check.

We are moving from AI experimentation to AI accountability. And in that world, context is not optional. It is the system that makes everything else work.

Best,

Ravit Jain

Founder & Host of The Ravit Show