Gartner: Companies Can Cut Agentic AI Costs 60% by Fixing Data Semantics, Not Upgrading Models

Companies deploying agentic AI are wasting up to 60% of their spend on agents that hallucinate, introduce bias, and produce unreliable outputs. The cause, according to new Gartner research presented at the firm’s Data & Analytics Summit in London, is not flawed models. It is data that lacks semantic context.

Gartner predicts that organizations prioritizing semantics in AI-ready data will improve agentic AI accuracy by up to 80% and reduce costs by up to 60% by 2027, according to Fortune, which first reported on the research’s CFO implications.

The Semantic Gap

“Agentic AI outcomes depend on context including semantic representations of data,” Rita Sallam, distinguished VP analyst at Gartner, said at the summit, as reported by IT-Online. “Without context, a clear understanding of the specific relationships and rules within an organization’s data, AI agents cannot operate accurately.”

The core argument: traditional schema-based data models organize information technically but fail to encode business meaning. An agent querying a database can pull the right rows but misinterpret what those rows represent in context, leading to hallucinated conclusions or biased outputs. Sallam warned that organizations failing to adopt comprehensive context structures “will perpetuate data inefficiencies and face heightened financial costs, as well as legal and reputational damage.”

Gartner’s proposed fix is a dedicated semantic “context layer” sitting at the core of enterprise data infrastructure, one that maps relationships, business rules, and organizational meaning on top of raw data. This is not a new database product. It is a structural requirement that Gartner argues should be treated as foundational, not optional.

A Capital Allocation Problem, Not a Technology Debate

Fortune’s Sheryl Estrada frames the finding as a capital allocation issue for CFOs: the AI conversation needs to shift from “which model should we buy?” to “is our data fit for purpose?”

Gartner expects regulators to demand greater semantic transparency, and boards to treat semantic governance as both a strategic risk and a competitive opportunity. “Context with semantic coherence will become a cost-control and trust strategy, not a nice-to-have,” Sallam said.

The timing is notable. According to FactSet analysis reported by Fortune, 65% of S&P 500 earnings calls in Q1 2026 cited “AI,” the second-highest share in five years. Companies are spending heavily. Whether they are spending well is a different question.

The Cancellation Wave

This research arrives alongside a separate Gartner prediction that over 40% of agentic AI projects will be canceled by end of 2027 due to escalating costs, unclear business value, or inadequate risk controls, according to MarTech. The two findings reinforce each other: companies deploying agents on poorly structured data face both higher costs and lower accuracy, precisely the conditions that trigger project cancellations.

DQ Channels notes that Gartner expects multistructured data management spending to account for 40% of total data management technology budgets by 2027, signaling that the shift Gartner advocates will require real capital commitment, not just policy changes.

The Data Readiness Gap

The practical implication for teams building or deploying agent systems: model selection matters less than data preparation. A smaller model with well-structured semantic context will outperform a frontier model operating on raw, uncontextualized data, at a fraction of the cost.

For enterprises that have already deployed agents, the research suggests an audit. How much of the current agent spend is going toward compensating for data problems that could be solved structurally? Gartner’s answer: potentially more than half.