Only 26% of enterprises have operationalized AI at scale, FPT-Forrester study finds

More than half of global enterprises are putting real money into artificial intelligence, yet only one in four has actually scaled it. That gap is the central finding of a Forrester Consulting study commissioned by FPT Corporation and released today, based on a survey of 397 business and technology decision-makers across North America, Europe, Asia Pacific, and Japan.

Investment is running ahead of execution

According to the research, 51% of organizations are allocating at least 5% of their IT budgets to AI. The ambition is clear. The follow-through is not: only 26% of respondents said they consider their organization advanced in operationalizing AI, as reported by Business Wire.

The study spans industries including automotive, financial services, healthcare, manufacturing, energy, and sports. Across all of them, automation and cost reduction are the primary drivers of AI deployment. Far fewer organizations are doing the harder work of redesigning business models around AI: just 34% reported pursuing an AI-first operating model.

Structural barriers, not technology gaps, are the real bottleneck

When enterprises hit a ceiling on AI scaling, the culprits are organizational as much as technical. The Forrester research identifies integration complexity, cited by 41% of respondents, and data silos, cited by 38%, as the top barriers to operationalizing AI. These are chronic infrastructure problems that no single model or tool can patch.

The governance picture is equally uneven. Only 39% of enterprises reported meaningful progress in aligning AI strategy, governance, and operating models. That misalignment limits the ability to scale AI in a controlled, consistent way across business units.

Measurement is arguably the sharpest problem. Thirty-five percent of organizations do not collect any quantified AI metrics, and 10% track AI outcomes in no form at all. Without that data, decisions about which pilots to scale and which to cut are essentially guesswork.

What enterprises want from AI partners

The study also mapped what procurement and technology leaders are prioritizing when selecting AI partners. The top three criteria are nearly tied: the ability to engineer, deploy, and operate AI across the full lifecycle (48%); strong governance and security capabilities (48%); and the ability to integrate with existing systems (47%).

Regional differences are notable for sourcing teams. North American and EMEA organizations placed the heaviest emphasis on full lifecycle capabilities, at 59% and 54% respectively. Organizations in Asia Pacific and Japan leaned more toward end-to-end strategic and execution support, signaling different partner expectations across geographies.

FPT's response: a platform and a framework

FPT is releasing the study alongside two specific offerings. The first is FleziPT, an AI platform backed by a global team of more than 30,000 AI-augmented engineers and AI Factories in Vietnam and Japan. FPT says the platform delivers up to 60% optimized development time, more than 50% less rework, and a 30% uplift in developer productivity, per Business Wire.

The second is FPT CASAN, a five-level AI transformation framework covering Curious, Augmented, Standard, Automatic, and Native stages. The framework is intended to give enterprises a structured way to assess AI readiness, tighten governance, and move AI out of isolated pilots into core business functions.

As AI ecosystems become more complex, organizations can no longer move forward in silos. They need partners who can bridge strategy, integration, governance, and operations to turn innovation into repeatable, enterprise-wide execution., Pham Minh Tuan, Executive Vice President, FPT Corporation and CEO, FPT Software

What this means for your team

• Audit your AI measurement practices first. If your team sits in the 35% that collects no quantified metrics, decisions about scaling or sunsetting pilots have no factual foundation. Define outcome KPIs before greenlighting the next initiative.
• Treat integration complexity as a procurement criterion. The study's top-cited barrier is not model quality but the ability to connect AI to existing systems. Vendor evaluations should weight integration depth and lifecycle support at least as heavily as model capabilities.
• Evaluate partners against full lifecycle criteria. Nearly half of surveyed enterprises are now prioritizing vendors who can engineer, deploy, and operate AI systems end-to-end. Assess whether current partners cover governance and operations, not just delivery.
• Use a maturity framework to benchmark readiness. FPT's CASAN and similar frameworks offer a structured lens for identifying where governance and operating model gaps are stalling scale, which helps align leadership before committing to the next phase of investment.