Confluent research has found that nearly three-quarters of global IT leaders say weak real-time data infrastructure is stalling efforts to scale artificial intelligence, suggesting that the underlying systems supporting AI are now a major constraint on deployment.
The company’s 2026 Data Streaming Report, based on responses from 4,625 IT leaders worldwide, found that 72% say a lack of real-time data infrastructure is slowing their attempts to scale AI. The same proportion reported encountering at least three challenges when scaling AI initiatives.
The most common barriers included insufficient infrastructure for real-time data processing, cited by 72%, uncertainty around data lineage, timeliness, and quality, cited by 66%, and fragmented data ownership, cited by 65%. Two-thirds of IT leaders also cited data infrastructure and data quality issues as barriers to agentic AI adoption, while only 32% reported having agentic AI in production.
Shaun Clowes, Chief Product Officer at Confluent, said: “Most organisations do not have an AI investment problem, they have a data problem. AI systems depend on fresh, accurate and contextual information, but too many are still being built on fragmented data, batch processes, and infrastructure that was not designed for continuous intelligence.”
The findings add technical depth to the AI adoption debate. Workforce skills have already been placed at the centre of deployment in Ministers tie AI adoption to workforce skills. Confluent’s research shows that trained teams and strong investment may still struggle if enterprise data is fragmented, stale, poorly governed, or difficult to access in real time.
AI projects often begin with models, tools, and use cases. Scaling them requires more reliable foundations. Customer service automation, fraud detection, supply forecasting, predictive maintenance, real-time pricing, operational copilots, and agentic workflows all depend on timely and trusted data. If data is delayed, duplicated, inconsistent, or owned by disconnected teams, AI systems may produce weak outputs or fail to integrate into critical processes.
The agentic AI finding is especially important. Agentic systems are expected to act with greater autonomy, executing tasks across workflows rather than only generating text or analysis. That creates a need for live context, permissions, auditability, and accurate signals from business systems. An agent acting on outdated data can create operational risk. An agent without clear lineage and governance can create compliance and accountability problems.
Confluent found that 80% of IT leaders say using enterprise data to drive AI-based systems is a top business priority. It also found that 88% say data streaming platforms help unblock agentic AI progress by making data more trustworthy, contextualised, and discoverable, while 94% say data streaming increases or is expected to increase the impact of AI investments.
Investment priorities appear to be shifting as companies confront those constraints. Data streaming ranked as a key priority for 88% of IT leaders, ahead of AI and machine learning technologies at 82%. That does not signal lower interest in AI. It shows growing recognition that model capability is only one part of the value chain.
Legacy estates will feel the pressure most sharply. Many enterprises still run batch processes, siloed databases, manual reconciliations, ageing integration layers, and inconsistent ownership models. AI can expose those weaknesses quickly. A proof of concept can work with a curated dataset; live deployment has to operate across messy workflows, security constraints, and customer-facing systems.
The management challenge is organisational as well as technical. Data ownership, governance, architecture, security, and accountability need alignment before AI can scale safely. IT teams cannot fix fragmented data alone when business units protect local systems or lack incentives to standardise.
AI’s next phase will reward companies that treat data infrastructure as a strategic asset rather than a back-office upgrade. Confluent’s findings indicate that the limiting factor for many organisations is no longer belief in AI, but whether their systems can support it at production speed.
You must be logged in to post a comment.