A new industry report has put a number on something many data leaders already suspect: most businesses have started deploying AI agents before their data infrastructure is anywhere near ready to support them.

Fivetran’s 2026 Agentic AI Readiness Index, published May 5 and based on a survey of 400 data professionals across the US, UK, EMEA, and Asia-Pacific, found that only 15% of organisations are fully prepared to support agentic AI in production. At the same time, nearly 60% are already investing millions in the technology — and 41% have already deployed agents in live environments.

That gap — between investment, deployment, and actual readiness — is the defining story of enterprise AI right now.

The Problem: Agents Are Making Decisions on Bad Data

The core issue isn’t the AI models. According to Fivetran CEO George Fraser: “Most companies aren’t failing at AI because of the models, they’re failing because their data isn’t ready. Organisations are pushing agentic AI into production on top of brittle pipelines, missing lineage, and systems that were never designed for autonomy. When that happens, you don’t get better outcomes — you get faster failures.”

AI agents are different from traditional software. A conventional dashboard or report surfaces information for a human to interpret. An agent takes action. It reads from your systems, reasons over that data, and makes decisions — often autonomously and at scale. If the underlying data is stale, incomplete, or ungoverned, the agent’s decisions will be too. And those decisions may already be running in your business.

The top three barriers companies cited for achieving their agentic AI goals were:

• Data quality and lineage (42%) — incomplete, inconsistent, or untraceable data
• Regulatory compliance and data sovereignty (39%) — especially acute for businesses operating across borders
• Security and privacy risk (39%) — agents touching sensitive systems without adequate controls

These are not edge-case problems. They are the centre of the enterprise AI challenge in 2026.

What Prepared Organisations Do Differently

The 15% that are fully prepared aren’t just more confident — they operate differently. Fivetran’s research found that ready organisations are significantly more likely to:

• Run always-on, automated data pipelines that keep information fresh and reliable in real time
• Enforce end-to-end data lineage and governance so every decision an agent makes can be traced back to a source
• Standardise on interoperable architectures that let data move freely across their infrastructure

As a result, these organisations can deploy agents more broadly — across internal workflows and customer-facing products — and are far more confident in their ability to generate real ROI from their AI investments.

The organisations that are struggling tend to have done the opposite: they’ve invested in the AI layer (the models, the orchestration frameworks, the front-end interfaces) but underinvested in the data layer underneath. Agents deployed on top of disconnected, poorly governed data are essentially making decisions in the dark.

What This Means for Business

This report matters because it names a problem that’s been quietly building across the industry. Boards and executives have been under pressure to show AI progress. That pressure has led many organisations to start deploying agents before the foundations are in place — and the results are starting to show up.

A few practical implications:

If you’re still planning your AI rollout, your data readiness is not a secondary concern. It’s the primary one. An AI agent is only as good as the data it can access, trust, and act on. Getting your pipelines automated, your lineage documented, and your governance policies in place before deployment will save you significant pain.

If you’re already in production, it’s worth an honest audit. Are your agents making decisions on real-time, trusted data — or are they drawing from systems that haven’t been updated since your last manual export? The gap between those two scenarios can be the difference between AI that creates value and AI that creates liability.

If you’re in data leadership, this report is a useful lever. The business wants agents deployed. That pressure isn’t going away. But the research now shows clearly that readiness is the variable that separates outcomes. You have the evidence to push for the data infrastructure investment needed to make agents actually work.

For businesses that already have strong data practices — clean pipelines, governed warehouses, real-time integrations — this is a genuine competitive advantage. The organisations that invested in data literacy and infrastructure over the past few years are not just better informed. They’re better positioned to extract value from AI agents while others are still debugging their foundations.

At Enterprise DNA, this is exactly what we’ve been building toward. A 220,000-person community of data professionals didn’t happen by accident. It happened because data foundations matter — and 2026 is proving that more clearly than any year before it.

---

The 2026 Agentic AI Readiness Index was conducted by Redpoint Ventures on behalf of Fivetran, surveying 400 data professionals across mid-sized and large enterprises in the US, UK, EMEA, and Asia-Pacific. Full report available at fivetran.com.

Want to build the data skills your team needs to make AI agents actually work? Explore Enterprise DNA’s learning platform — built for business teams who need to move from data chaos to data confidence.