Only 26% of enterprises say they have successfully moved artificial intelligence from pilots into live operations, according to a global study released July 8 by Forrester Consulting and commissioned by FPT Corporation. The findings point to a widening execution gap: spending on AI is rising, but business outcomes are not keeping pace.
Budgets grow while AI remains stuck in pilot mode
The research, based on surveys and interviews with 397 technology and business leaders across North America, Europe, Asia Pacific and Japan, shows more than half of organizations—51%—now allocate at least 5% of their IT budgets to AI. Automation and cost reduction top the list of objectives. Yet only 34% report pursuing what the study defines as an AI-first operating model. In practical terms, most companies are applying AI tactically to discrete use cases rather than reshaping core processes.
The report, titled From Pilots to Reusable Platforms: A Blueprint for Scaling Enterprise AI, spans industries including automotive, financial services, healthcare, manufacturing, energy and sports. Its central message: the hard part of scaling AI is no longer tooling—it is the operating model, data foundations and cross-team execution needed to sustain production-grade systems.
Measurement blind spots undercut decision-making
Forrester highlights a pervasive lack of measurement discipline. 35% of organizations do not collect quantified metrics for AI at all, and 10% track no AI outcomes in any form. Without outcomes data, leaders have little evidence to decide which pilots merit further investment, which should be retired and where platform capabilities should be standardized for reuse.
| Metric | Share of respondents |
|---|---|
| AI operationalized at scale | 26% |
| Allocate ≥5% of IT budget to AI | 51% |
| Pursuing an AI-first operating model | 34% |
| No quantified AI metrics collected | 35% |
| No AI outcomes tracked at all | 10% |
Integration and data remain the top blockers
When asked why proofs of concept fail to become platforms, two long-running IT issues dominate responses, now magnified by AI’s data hunger and real-time needs:
- Integration complexity cited by 41%
- Data silos cited by 38%
AI systems depend on consistent, governed data flows across applications that were never designed to interoperate. Without modern integration patterns and shared data models, organizations struggle to deliver reliable inputs to models, to monitor performance and to embed outputs in everyday workflows.
What it means for Canadian enterprises
Although global in scope, the study includes respondents from North America and reflects challenges common across Canadian industry: strong ambitions and budget commitments, tempered by practical obstacles in data readiness, integration and operating model change. The headline number—only one in four organizations have AI running at scale—suggests many teams here are still consolidating pilots, building shared services and agreeing on what success looks like.
For leaders pushing beyond experiments, the report’s numbers point to three immediate priorities: establish consistent outcome metrics, tackle integration as a product (not a project) and treat data governance as a prerequisite to any model deployment. That approach aligns budgets with results and helps convert one-off wins into reusable capabilities.
Methodology and context
The study was conducted by Forrester Consulting and commissioned by FPT Corporation. It draws on input from 397 decision-makers across multiple sectors and regions. As with any commissioned research, readers should consider the sponsor’s interests alongside the methodology. Still, the cross-industry patterns it surfaces—rising investment, low operationalization rates, and persistent data and integration challenges—mirror what many technology teams report anecdotally as they try to scale AI from proof of concept to production.
