AI's Latest Hurdle: The Data Dilemma

Enterprises are rushing to harness AI, but outdated data systems hold them back. While AI shows promise, only 10% of companies are scaling their efforts. Trust and context issues plague progress.
AI is the new darling, the shiny tool that promises to revolutionize everything from supply chains to financial planning. But despite the rush to adopt AI, most companies can't scale their AI efforts. Only 10% have managed to make their AI agents truly impactful. Why? Because their data infrastructure is as outdated as a rotary phone.
The Data Roadblock
In late 2025, McKinsey found that nearly two-thirds of companies were dabbling with AI agents, and 88% had AI integrated into at least one part of their business. But let's not pop the champagne just yet. The real party pooper here's data. Irfan Khan of SAP Data & Analytics reminds us that AI agents are only as good as the data they munch on. And most companies don't have the gourmet data architecture needed to serve them a proper meal.
The lack of a strong data foundation isn't just a minor hiccup. It’s the elephant in the room. Companies need to stop stockpiling data like doomsday preppers and start ensuring it has the right business context. Without that, AI is like a high-performance car stuck in gridlock traffic.
Trust Issues and Data Sprawl
Trust in data is at an all-time low, with two-thirds of business leaders admitting they don't trust their own data. This 'trust debt' is slowing down AI adoption faster than you can say 'data breach'. Semantic consistency and reliable operational context are the antidotes, but who has time to focus on that when there's flashy new tech to play with?
To make matters worse, data sprawl is rampant. Data is scattered across countless clouds, lakes, and warehouses, thanks to the architectural shift of separating compute from storage. More than two-thirds of companies grapple with data siloes, and over half are juggling over a thousand data sources. Spare me the roadmap, techies. It’s a mess.
AI's Role in the Software Stack
Some dreamers think AI agents will make SaaS applications obsolete. Khan says, 'Not so fast.' SaaS isn't going anywhere. AI agents will be another layer in the software stack, not the stack itself. Companies won't just toss away their general ledgers for AI agents. What are these agents going to do without business context and processes? Perform interpretive dance?
As we look to the future, the real challenge is making AI agents first-class citizens in business processes. That won't happen without a semantic or business-aware layer to make sense of the chaos. But let’s face it, most companies are as prepared for this as they're for a zombie apocalypse.
For enterprises wrestling with AI deployment, the message is clear: focus on the data that matters, invest in governance and semantics, and resist the urge to automate everything too soon. Rushing into AI without a solid data foundation is a surefire way to end up with a high-tech mess on your hands.
Get AI news in your inbox
Daily digest of what matters in AI.