The News
Plug and Play released a pulse survey of Fortune 500 and Forbes Global 2000 companies, finding that 74% of large enterprises now have at least one AI solution in production. Despite that adoption breadth, half of respondents report they are either too early to assess ROI or are not consistently measuring it. The survey identifies data foundations as the top barrier to scaling AI, cited by 71% of respondents, followed by governance friction at 53% and legacy system integration at 26%.
Analyst Take
Enterprise AI has cleared the adoption threshold. Three-quarters of the world’s largest organizations are running AI in production. The question that defined 2024 and 2025 (“Are you doing AI?”) has been replaced by a harder one: “Can you prove it’s working?” That the majority of enterprises cannot yet answer that question consistently is not a sign of failure. It’s a sign of where the real work begins.
The Data Problem Isn’t a Surprise, But Its Scale Is
Seventy-one percent of respondents flagging data foundations as the top barrier to scaling AI is a striking number, but it shouldn’t shock anyone who has watched enterprise software cycles over the past two decades. The pattern is familiar. Organizations adopt a new technology category faster than they can build the underlying infrastructure to support it. What’s different with AI is the consequence. A poorly tuned CRM slows sales. A poorly grounded AI model generates confident, wrong outputs at scale, and those outputs increasingly touch customers, financial models, and operational decisions.
For IT decision-makers, this is a capital allocation signal. Investment in data quality, pipeline integrity, and governance infrastructure is not preparatory spending that comes before the AI budget. It is the AI budget. Organizations that treat data and governance as separate line items from AI will continue to struggle to measure, and therefore justify, the returns on their AI programs.
What “In Production” Actually Means Here
The survey’s segmentation of production AI deployments is worth examining closely. Thirty-seven percent of enterprises running AI in production do so within a single business function. Only 32% operate it across several functions. Just 5% consider themselves AI-native. This distribution tells a story of concentrated, contained deployments rather than enterprise-wide transformation. Running a single AI tool in one department is production, technically, but it is not scale.
For developers and architects, this containment pattern matters architecturally. Single-function deployments rarely stress-test the infrastructure decisions that become critical at scale. Model versioning, prompt management, inference cost control, multi-team access patterns. ECI Research’s 2026 Application Development survey found that 65.2% of respondents selected “0–20” when asked what percentage of engineering time is spent on net-new innovation, suggesting that most engineering organizations are already capacity-constrained. Expanding AI from one business function to many without purpose-built platform infrastructure will consume that innovation budget fast.
The Governance Gap as Competitive Differentiator
Governance friction at 53% is the finding that deserves the most attention from both audiences. Governance is the mechanism through which AI ROI gets measured, controlled, and defended in front of boards and regulators. Organizations that cannot measure AI value are, in most cases, organizations that have not yet built the governance layer that makes measurement possible. This isn’t philosophical. It’s operational.
ECI Research’s 2026 Application Development: Day 1 survey found that 58.2% of respondents selected “Moderate increase (10–25%)” when asked how much they will increase AI governance spending. That signals a market that already understands the gap. The companies that close it fastest will have a durable advantage. Not just better AI outputs, but auditable, explainable, and cost-attributable AI outputs that can survive procurement scrutiny, regulatory review, and executive skepticism.
The build-versus-buy dimension mentioned in the report adds another layer. Enterprises evaluating AI vendors increasingly need to ask not just “Does this model perform well?” but “Does this vendor’s architecture support the governance and observability capabilities we need to prove ROI internally?” That question will reshape vendor selection criteria in the next 12–18 months.
Looking Ahead
The next phase of enterprise AI competition will not be fought on model capability. It will be fought on measurement infrastructure. The organizations that can instrument their AI deployments, attribute cost and outcome at the function level, and demonstrate ROI with the same rigor applied to any other enterprise software investment will pull ahead. Those that cannot will face growing internal pressure to consolidate or cut AI programs, regardless of the underlying potential of the technology.
Plug and Play’s position as a corporate innovation intermediary gives this survey a particular signal value. When the firms advising Fortune 500 companies on AI adoption are saying that half of those companies can’t measure what they’re getting, that’s a market condition, not a temporary lag. Expect data platform vendors, AI governance tooling providers, and observability platforms to sharpen their enterprise AI positioning aggressively over the next two to three quarters. The ROI measurement problem is real, it’s broad, and it’s solvable. The vendors who credibly solve it will find a very receptive buyer.
Stay Ahead of Application Development Trends
Get weekly analyst insights, research notes, event coverage, and AppDevANGLE updates delivered directly to your inbox.
Subscribe for Weekly Insights
Join technology leaders, practitioners, and GTM teams following the trends shaping modern software delivery.
Looking for deeper research access?
Explore ECI Research reports, survey insights, and market analysis through the ECI Research Portal.
