New Intelligence is Moving Faster than Enterprise Controls
AI is being integrated into core enterprise systems faster than many organizations can secure and govern it. A new global study from NTT shows companies expanding AI deployment while gaps in infrastructure readiness, data integrity controls, and governance frameworks continue to limit safe operation at scale.
Investment Continues Despite Uneven Outcomes
AI spending continues to rise across regions and industries. AI is viewed as central to competitiveness and long-term strategy, which keeps budgets growing even when outcomes vary. About half of organizations say current AI initiatives meet expectations, while the remainder report weaker returns.
These differences do not stem from a lack of interest. Infrastructure limits emerge as the primary constraint. Systems designed for earlier workloads strain under large models, frequent retraining, and data-heavy pipelines. These pressures slow deployment and increase operational complexity as usage expands.
“Amidst historic AI investments, business leaders are asking vital questions about achieving an AI-empowered future with the potential to unlock unprecedented growth and productivity but without sacrificing quality, resilience, and the employee and social contract that is expected of businesses,” said the CEO and Chief AI Officer of a leading company.
Infrastructure Readiness Lags Behind Demand
Only a small share of companies say their infrastructure can support AI at scale. Most remain in transition, adapting legacy systems or introducing new components alongside older platforms. Shortages in computing capacity, network throughput, and data preparation appear repeatedly across responses.
Development cycles lengthen, production releases slow, and AI remains harder to operationalize across teams. Programs continue to advance, while friction builds as workloads grow.
The study frames infrastructure as a strategic concern. Treating compute, networks, and data pipelines as long-term assets reduces bottlenecks as AI use expands.
Performance Shapes Early AI Design Choices
Performance drives most AI infrastructure decisions. Organizations focus first on meeting model size, latency, and reliability requirements. Energy use and environmental impact often receive attention later in the lifecycle.
Concern about energy implications remains common during early planning. A large share of respondents believe sustainability efforts reduce profitability, which shapes how infrastructure investments are sequenced.
The research shows a growing alignment between performance and energy outcomes. Distributed architectures, advanced cooling, and optical networking appear frequently in plans to support expanding AI workloads.
Photonics Gains Attention as Workloads Grow
Photonics stands out as one of the most recognized infrastructure technologies in the study. Respondents associate it with higher throughput and lower energy demands, qualities that align with AI-intensive environments. Interest increases with company size, where data movement and heat management demand greater control.
Integration complexity, upfront costs, and uncertainty around returns slow deployment. Many organizations place photonics on a medium-term evaluation path rather than immediate rollout.
The study presents photonics as part of a broader search for infrastructure that can absorb AI growth without proportional increases in strain on power and hardware.
Data Integrity Defines Trust in AI Systems
AI performance is closely linked to the quality and integrity of underlying data. Respondents say their organizations need to do more to clean, safeguard, and govern the data feeding AI systems.
Weak data hygiene introduces risk. Poor inputs lead to unreliable outputs, weaker decision support, and greater exposure to security incidents. These risks increase as AI systems move from pilots into core workflows.
Widespread use of unsanctioned AI tools introduces new risks across enterprises. Sensitive data leakage, erosion of data integrity, and security vulnerabilities rank highest among concerns. Inaccurate outputs also stand out, particularly when AI tools influence business decisions without oversight.
The study identifies shadow AI as a systemic issue tied to easy access to public AI tools and pressure to move faster.
Governance Maturity Differs Across Organizations
Formal AI governance exists in many organizations, though confidence in its maturity varies. Some report structured oversight through governance councils, risk assessments, and access controls. Others acknowledge gaps between policy and day-to-day practice.
Interest in agentic AI systems intensifies these concerns. Autonomous decision-making increases the impact of governance weaknesses, with cybersecurity and data protection cited as leading risks.
Responses describe layered controls that include restricted environments for sensitive workloads, privacy-enhancing techniques, and tighter role-based access. Governance increasingly spans planning, deployment, and operation.
