Share with your CDO
The “SaaSpocalypse” that briefly wiped $300 billion from Salesforce, ServiceNow, and Adobe in February turned out to be a misdiagnosis, not a prediction. The iShares enterprise SaaS ETF (IGV) dropped 30% by April and recovered to near flat by mid-September, which tells you the market overcorrected on the thesis that AI agents would kill the app layer. What’s actually happening, per this Reltio-sponsored analysis in Business Insider, is that systems of record are becoming infrastructure, and the data quality living inside them is now the real competitive variable.
What this means for your business
The organizations most exposed here aren’t the ones still running Salesforce or Workday. They’re the ones planning to replace those systems because they misread the threat. If 94% of companies are exploring or implementing AI but only 15% consider their data foundation ready for agentic AI, the bottleneck isn’t the application and it isn’t the model. It’s the records those agents will act on, not just read. Bad data in a chat interface produces a wrong answer. Bad data feeding an autonomous agent produces a wrong action, and sometimes one that can’t be reversed.
The HBR Analytic Services survey cited here, sponsored by Reltio (which sells a data unification platform, giving it an obvious incentive to frame data quality as the decisive constraint rather than model capability or agent orchestration), still lands on a number that should disturb any data leader: 94% of executives rank data trustworthiness as their most critical capability, but only 39% say they’re actually proficient at it. That gap between stated priority and demonstrated capability is where agentic AI projects die. Gartner’s prediction that 40% of agentic AI initiatives will be canceled by end of 2027 is consistent with the MIT NANDA finding that 95% of generative AI pilots have produced no measurable financial return so far.
The shift from application-layer to data-layer competition reframes one decision most CDOs already own: whether the data governance program is scoped as a compliance function or as the architecture on which the next generation of automation runs. Those are genuinely different mandates, with different staffing, tooling, and reporting implications. If agentic AI is on the roadmap and the data unification work isn’t funded alongside it, the roadmap is fiction. I’d revise that view if enterprises start shipping agents at scale on messy, siloed data and the error rates prove manageable, but Informatica’s finding that half of companies already using agentic AI cite data quality as their single biggest production barrier suggests that bar hasn’t been cleared.
Concept deep-dive: Agentic AI
Agentic AI refers to AI systems that don’t just answer questions but take sequences of actions autonomously, querying databases, drafting documents, updating records, or triggering workflows, to complete a goal with minimal human intervention per step. Think of it as the difference between a GPS giving you directions and a self-driving car executing them. The business implication is that errors compound across steps rather than stopping at a single wrong output, which is why the quality of the underlying data becomes load-bearing in a way it never was for read-only AI tools.
Based on reporting from AI Agents Reshape Enterprise Software Without Obsoleting It, originally published 2026-10-01 13:11:00.
