Data Foundation for AI Success: Key Takeaways from Gartner's Summit

The opening keynote at Gartner’s Data & Analytics Summit in London started with a sobering stat: only 1 in 5 AI initiatives deliver ROI.

The reasons weren’t what you’d expect from an AI conference. Not model quality. Not compute. Not the state of foundation models. The culprits named were lack of understanding of costs, inability to scale, and the one that landed loudest in that room was data quality.

For anyone who’s spent time in the enterprise data space, this isn’t a new finding. But hearing it delivered as the opening salvo of Gartner’s flagship European event, to an audience of CDOs, CDAOs, VP-level data leaders, and platform architects who have been living this problem for years, felt different. Gartner wasn’t diagnosing a fringe issue. They were naming the central blocker to AI value at enterprise scale and building an entire framework around what to do about it.

That framework is worth unpacking. Because it maps almost exactly to the sequence we’ve been arguing for.

What Gartner statistic reveals the importance of AI foundations?

The keynote reframed ROI, not as a financial metric, but as three compounding dimensions that have to be built in order:

• Return on Intelligence is about setting your AI ambition. Gartner laid out three organizational archetypes: AI-Cautious, AI-Opportunistic, and AI-First. The honest question isn’t which you aspire to, it’s which you actually are, based on your current data infrastructure and organizational readiness. The whitewater rafting metaphor they used stuck with me: every enterprise is already on the water. The question is whether you’re navigating it deliberately or getting swept by the current.
• Return on Integrity is where Gartner got specific about the foundation requirements. AI-ready data isn’t just governed data, it’s connected, rationalized, embedded into workflows, and contextualized. “Context is king” was stated plainly and without hedging. Semantic layers, ontologies, knowledge graphs, Gartner didn’t present these as advanced maturity concepts. They positioned them as prerequisites. The enterprise cases they cited to illustrate this point, a multinational industrial technology leader specializing in energy management and automation reducing compliance issues by roughly 20% through automated policy-as-code, a multinational, enterprise-level B2B supply chain orchestrator using automation to classify and govern data at scale, were all stories about operationalizing data trust before deploying AI, not alongside it.
• Return on Individuals was the third pillar, and the most human one. The framing wasn’t “AI will replace your team.” It was nearly the opposite: every employee will be valued increasingly based on skills, not role. Humans aren’t just in the loop, they’re in the lead. Mindset and skillset matter more than which tools you’ve licensed. It’s a point that’s easy to nod along to and hard to actually operationalize, but Gartner was emphatic that organizations treating AI as primarily a headcount story are misreading what’s happening.

The Stat That Should Make Every Data Leader Stop

Buried in the keynote was a data point that deserves more attention than conference slides typically get.

Gartner compared organizations most satisfied with their AI outcomes against those least satisfied and looked at where each group was spending. The most satisfied organizations were investing substantially more on AI foundations, data management, governance, and talent, than the least satisfied. The gap on spend toward AI technology itself was much smaller.

The implication is powerful: the organizations getting real ROI from AI aren’t the ones with the best models or the most GenAI pilots. They’re the ones that built the foundation first. They treated data infrastructure as a prerequisite, not a parallel workstream.

The least satisfied organizations, the ones struggling to get past pilot, struggling to scale, struggling to show value, were trying to build the intelligence layer before the integrity layer was solid. And they were paying for it.

This isn’t a novel insight if you’ve been thinking about this problem for a while. But Gartner putting empirical weight behind it changes the conversation in a useful way. It gives data leaders something concrete to bring into budget discussions and boardroom conversations about AI readiness.

How does Anomalo's platform align with Gartner's ROI framework?

Gartner’s case examples across the keynote reinforced the same pattern in different contexts.

A UK local authority, responsible for the administration of a massive, predominantly rural landmass with a highly dispersed population, built an AI agent to help surface the human consequences of data decisions, connecting what happens in datasets to what it means for real people in their community. The message Gartner drew from it was pointed: there are human consequences to our data decisions. Their success was the result of making the underlying data trustworthy and meaningful enough for an AI agent to act on responsibly.

A large, community-focused, private nonprofit health system serving Southwest Florida used trusted data infrastructure to transform their patient care quality ratings from one star to five. A premier multinational law firm built proactive systems to flag issues and negative feedback before they compounded downstream. Transport for London enabled data-driven operations at a scale that affects millions of daily journeys.

None of these are stories about AI outperforming human judgment. They’re stories about organizations that got serious about their data foundation, and then discovered that AI could do more, more reliably, once that foundation was in place. The sequence wasn’t optional. That was the point.

Why is the window for building an AI foundation shorter than it looks?

Sitting in that keynote, I kept finding myself thinking about how closely Gartner’s three-part framework tracks the three phases we’ve built the Anomalo platform around. It wasn’t a forced connection. The logic of the problem leads you to the same place.

Phase 1, Data Monitoring, is Gartner’s Return on Integrity made operational.
The “Strengthen Your Foundation” pillar calls for governing, connecting, and rationalizing your data. But most enterprises face a more elemental problem first: they can’t reliably tell you what changed in their data overnight, which tables are drifting, or where freshness has degraded. That’s not a governance gap. It’s a monitoring gap, and you can’t govern what you can’t see.

Continuous, ML-driven monitoring, the kind that learns what normal looks like for each table rather than requiring someone to write and maintain rules, is the practical foundation layer Gartner is describing. It’s what gives data teams the confidence to say that the data feeding their AI systems is what they think it is. Without it, the integrity pillar is more aspiration than infrastructure.

Phase 2, Data Understanding, is where Gartner’s “context is king” claim becomes concrete.
The keynote was unusually emphatic here. Translating technical data language into business language isn’t an advanced capability for mature organizations. It’s a requirement for AI-ready data. Semantic layers, ontologies, knowledge graphs, these aren’t nice-to-haves. They’re the context layer that makes autonomous agents useful rather than dangerous.

Anomalo surfaces meaningful changes in your data automatically, in plain language, before anyone has to ask. Documentation synthesizes context from Slack conversations, existing runbooks, wikis, and Anomalo’s own statistical understanding of your tables, and keeps it current without manual effort. What Gartner is calling for is a living, continuously updated context layer on your enterprise data. That’s the problem Phase 2 is built to solve.

Phase 3, Agentic Analytics, is Gartner’s Return on Intelligence, operationalized.
The AI-First archetype Gartner described isn’t about having the most AI pilots running. It’s about organizations where agents are doing real work, monitoring, investigating, triaging, surfacing, on a foundation that’s trustworthy enough to act on autonomously. That’s a different bar than “we have AI deployed.”

Anomalo is built toward that bar. Data issues get investigated and triaged before anyone is paged. Business metrics get watched continuously without anyone checking dashboards. Analytics get delivered proactively, not reactively. We call the category Self-Driving Data. Gartner would call it, in their framing, the AI-First archetype with a solid Return on Integrity underneath it. Both descriptions are pointing at the same thing.

The Window Is Shorter Than It Looks

Gartner’s framing for the broader moment was the shift from the Age of Digital to the Age of Intelligence, and their investment data suggests that transition is already well underway. Meaningful portions of CIOs surveyed are increasing their GenAI and AI investments, and analytics and BI spend is following. The direction of travel isn’t uncertain.

What Gartner was implicitly arguing, and what the enterprise case studies kept reinforcing, is that the organizations that build the foundation now will compound their advantage over time. Every agent interaction that teaches the system something about your data is institutional memory your competitors don’t have. Every dataset properly monitored and contextualized is a building block for more reliable autonomous operations.

The sequence matters. Foundation first. Context second. Agents third. Not because it’s a clean taxonomy, but because each phase genuinely enables the next. You can’t have trustworthy context without a trustworthy foundation. You can’t deploy reliable autonomous agents without both.

Gartner made the case in London for why this isn’t a “get to it eventually” problem. The gap between the organizations most and least satisfied with their AI outcomes is already visible in the data. It’s a gap built on choices about where to invest and in what order, and it’s widening.

If you’re thinking through where your data infrastructure sits on this journey or want to see how the Self-Driving Data platform maps to the foundation Gartner is describing we’d be glad to walk you through it in your environment.

Request a Demo Contact Us