Nearly 60% of Enterprises Stuck in AI Pilot Phase as Deployment Stalls Measurable ROI: Citi Report

Despite hundreds of billions of dollars poured into enterprise artificial intelligence over the past two years, the corporate world is encountering a steep implementation wall. According to a research report from Citi Institute, nearly 60% of surveyed organizations remain stuck in exploration or pilot stages, unable to push projects into full-scale operational production.

The findings highlight that while foundational models have become ubiquitous, embedding algorithms reliably into daily enterprise workflows remains the chief hurdle preventing businesses from realizing tangible balance-sheet value.

A Stark Regional Adoption Divide The transition from experimental sandboxes to enterprise production is fracturing along regional lines:

North America Surges Ahead: Roughly 77% of surveyed North American companies have pushed AI into live production or enterprise-wide scaling.

Asia-Pacific Stalls in Pilots: APAC organizations remain largely concentrated in exploratory phases, running low-stakes trial runs and internal upskilling programs.

Europe in the Middle: European enterprises occupy an intermediate position, balancing regulatory and data-governance guardrails against deployment urgency.

The Scaling Dividend: 15%+ Value Creation For the 40% of organizations that have breached the production barrier, the financial payoff is substantial:

• Quantifiable Gains: Sixty percent of companies operating in the production and scaling phases report transformational value creation exceeding 15%, realized through automated cost cuts, sustained employee productivity jumps, or direct top-line revenue expansion.

• Strategic Motivation: Transforming the core business model has become the primary driver for corporate AI capital expenditure (cited by 34% of respondents), followed by internal capability building (32%) and targeting specific operational performance benchmarks (25%).

"AI's biggest opportunity lies in redesigning business models, rather than focusing only on automating existing tasks." — Citi Institute Report

Moving Beyond Task Automation to Autonomous Agents While routine task automation remains the baseline entry point for most corporate programs, leading enterprises are shifting capital toward autonomous agents designed to manage multi-step, dynamic workflows across fragmented software stacks.

The report emphasizes that winning in the next deployment phase has little to do with simply buying access to off-the-shelf foundation models. Instead, sustainable enterprise differentiation depends on three unglamorous pillars: proprietary clean data, specialized internal engineering expertise, and tight integration into legacy core architectures. Furthermore, firms scaling AI are compressing their investment payback horizons—treating the technology as a core operating requirement rather than an open-ended R&D experiment.