Innovation & Tech

54% of Enterprises Killed Agentic AI Projects Over Unclear ROI

A SnapLogic-commissioned survey of 1,000 IT decision-makers finds 54% of enterprises abandoned AI agent projects, even as 85% say they're confident scaling AI adoption.

By Amara Osei

5 min read

Research Finds 54% of Enterprises Abandoned Agentic AI Projects Over Unclear Business Value
Research Finds 54% of Enterprises Abandoned Agentic AI Projects Over Unclear Business ValueAI-generated

What's News

  • 54% of surveyed IT leaders said their organizations abandoned AI agent projects because business value was unclear, per SnapLogic's 2026 Agentic Enterprise Readiness Index.
  • 85% of respondents expressed confidence in scaling AI adoption, and 93% reported having comprehensive AI-agent policies in place.
  • Gartner projects over 40% of agentic AI projects will be canceled by end of 2027 due to costs and unclear value.
  • CloudZero finds reported AI spending runs roughly 12x lower than actual AI-driven cloud consumption.

More than half of enterprises — 54% — have abandoned agentic AI projects because they couldn't demonstrate clear business value, according to a survey of 1,000 IT decision-makers commissioned by integration-platform vendor SnapLogic and released November 17.

The finding anchors the 2026 Agentic Enterprise Readiness Index, conducted by research firm 3GEM between July 21 and August 4 among IT leaders at enterprises in the United States, United Kingdom, and Germany. It lands amid record AI budgets and a growing pile of evidence that spending and measurable returns have decoupled.

Confidence high, results thin

The survey's readiness indicators read strong. Eighty-five percent of respondents said they were extremely or very confident in their ability to scale AI adoption. Ninety-three percent reported a comprehensive AI-agent policy or framework in place, or one covering most of the business. Ninety percent said they trusted, or would trust, outputs produced by AI agents.

Yet half felt pressure to keep investing in AI agents even when business value proved hard to demonstrate, and 48% said they struggle, or expect to struggle, to measure ROI from agentic initiatives.

SnapLogic CTO Jeremiah Stone framed the gap as a maturation problem: "Organizations are putting the governance, infrastructure, and investment in place to scale AI agents, but those foundations alone do not guarantee business outcomes."

Overspend is the norm

Cost visibility hasn't produced cost control. Eighty-nine percent of respondents reported some level of AI-agent overspend, with nearly two-thirds calling it moderate or significant. While 88% said budget responsibility was clear and 81% said they could accurately measure the cost of individual agents, only 42% said they could accurately forecast future AI and token spending. Thirty-six percent said AI investment was outpacing their ability to govern and operationalize agents.

The survey scores organizations across five dimensions SnapLogic defines: Agent Readiness, Integration & Orchestration Readiness, Governance & Trust Readiness, Economic Readiness, and Value Realization Readiness.

One caveat deserves weight. SnapLogic sells what it calls an "AI operational layer for the Agentic Enterprise" — an integration and governance platform for connecting applications, data, and AI agents. The company commissioned the survey and announced the results at its own INTEGRATEAI 2026 conference in London. Vendor research is evidence of what the vendor claims, not independent verification.

What do outside numbers show?

Independent data points the same direction. Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, up 47% year over year, while McKinsey finds only 37% of organizations attribute any EBIT impact to AI — flat from 2025. S&P Global reported 42% of companies abandoned most of their AI initiatives in 2025, up sharply from 17% a year earlier. RAND estimated over 80% of AI projects fail to deliver intended business value, roughly twice the failure rate of comparable non-AI IT projects.

The forward projections are just as blunt. Gartner expects at least 50% of generative AI projects to overrun budgets through 2028, and over 40% of agentic AI projects to be canceled by the end of 2027 due to escalating costs, unclear business value, and inadequate risk controls. Flexera puts wasted cloud spend at 29%, reversing a five-year improvement trend it attributes to surging AI workloads. Average GPU utilization sits at 23%.

KPMG finds only 11% of organizations have reached the stage of deploying and scaling AI agents in ways that produce enterprise-wide outcomes. Meanwhile, 90% of CEOs believe AI agents will enable measurable ROI in 2026, and they've committed more than 30% of 2026 AI investment to agentic AI. (These figures appear in CloudZero's 2026 AI statistics compilation, which cites the original firms.)

The accountability gap

CloudZero, a cloud-cost intelligence vendor, calls the dynamic an "AI accountability gap": the share of AI spend no one can trace to a product, feature, customer, or outcome. Its research finds reported AI spending runs roughly 12x lower than actual AI-driven cloud consumption. Organizations budget 30–36% of cloud spend for AI, yet AI-specific line items show up at just 2.5% of bills. Fifty-seven percent still track AI costs in spreadsheets; 15% have no formal AI cost-tracking system at all.

SnapLogic used the survey's launch to position its own roadmap as the fix. This week it announced an MCP Discovery Hub for finding and reusing Model Context Protocol servers, enhanced pipeline governance in SnapGPT, and planned Agent Observability capabilities. It also announced a partnership with Honeycomb for cross-system observability of AI agents.

What the data can't answer

The press release doesn't disclose how many abandoned projects were pilots versus production deployments, and it offers no breakdown by industry, company size, or use case. It never defines "unclear business value" — whether projects failed to set success metrics, failed to hit them, or lacked the instrumentation to measure them at all. And the sample consists exclusively of IT decision-makers; business-unit sponsors, finance leaders, and end users weren't surveyed.

The pattern the survey captures is real: enterprises have built the organizational muscle to deploy agents at scale — policies, data access, trust — but lack the instrumentation to connect agent activity to business outcomes. Whether the 54% figure reflects failed experiments that should have died sooner, or promising initiatives starved of measurement, the data can't say.

What's clear is where the judgment now falls. The next phase of enterprise AI adoption will be measured not on deployment velocity but on attributable value — and the vendors and internal teams that link every token, GPU hour, and agent run to a measurable outcome will set the standard for the rest of the market.

By Amara Osei

Source: epyhia.com

Share this article:

More from Amara Osei

Amara Osei

Show full bio

Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.

2,169 articles

Related articles

« Previous articleNext article »