Many organisations have purchased AI tools yet still struggle to demonstrate returns. Cloudflare’s 2026 App Innovation Report argues that the missing ingredient often sits underneath the model: applications that can provide dependable data, flexible integration, and secure operations.
The report surveyed more than 2,300 senior leaders across APAC, EMEA, and the Americas. Organisations ahead of schedule on application modernisation were nearly three times more likely to report a clear payoff from AI investment. In APAC, 92 percent of leaders called software updates the single most important factor in improving AI capabilities.
The finding reframes AI performance as an architecture and operating-model problem, not simply a model-selection problem.
Modern Applications Let AI Join Real Work
Legacy systems make experimentation deceptively easy and scaling painfully difficult. A pilot may read a static data extract, while production requires reliable access controls, live interfaces, monitoring, and connections to brittle workflows.
Modernised applications create reusable integration points and reduce the rework required when AI moves from demonstration to everyday use. Cloudflare describes a reinforcing cycle: companies update applications to support AI, then use successful AI outcomes to justify deeper modernisation.
In APAC, 90 percent of leading organisations had already integrated AI into existing applications, and about 80 percent planned to increase that integration during the following year. The emphasis has shifted from testing a standalone capability to embedding assistance inside customer experiences, internal processes, content applications, and revenue-generating work.
That integration makes value measurable. Leaders can observe completion times, customer outcomes, reliability, and costs within an established process instead of relying on pilot anecdotes.
Security Alignment Determines Sustainable Speed
Organisations behind on modernisation often act after a security incident or operational failure. Their teams spend more time tracking application and API risks, repairing gaps, and managing technical debt. That reactive work reduces confidence and leaves less capacity for AI deployment.
Cloudflare finds that alignment between application and security teams is associated with stronger scaling. When security is part of design, teams can apply consistent controls as services change. When it remains a late-stage review, every integration becomes a negotiation and surprises arrive near release.
Reliability is therefore a practical speed limit. AI may generate an answer quickly, but the surrounding application still needs dependable identity, data access, observability, and recovery. An unstable foundation converts rapid inference into operational risk.
Modernisation does not mean replacing everything. Teams can prioritise the applications closest to valuable AI workflows, expose controlled interfaces, improve visibility, and remove the dependencies that cause repeated failure. The sequence should follow business outcomes rather than architectural fashion.
Fewer Tools Give Developers Room to Improve
Tool sprawl compounds the problem. Nearly every organisation in the APAC data reported difficulty managing large, complex technology stacks. Around 86 percent of leaders were actively removing redundant tools and addressing shadow IT.
Consolidation can reduce cost, but its larger benefit is clarity. Fewer overlapping platforms make data flows easier to understand, security controls easier to apply, and AI integrations easier to maintain. Developers can improve working systems instead of repeatedly configuring replacements or rebuilding functions from scratch.
The report’s strongest implication is that AI and modernisation should not compete as separate budget lines. They are connected work. An AI initiative can reveal which application bottlenecks matter, while modernisation supplies the conditions that turn a promising output into a reliable service.
Leaders should still interpret survey associations carefully. Companies that modernise early may have stronger management, investment capacity, or engineering practices that also contribute to AI returns. The report does not prove that one upgrade will triple value.
It does, however, identify a credible pattern. Organisations that treat AI as an attachment to fragmented systems tend to produce shallow gains. Those that align applications, security, and integration create more places where AI can affect an actual outcome.
The practical question is no longer whether an enterprise has access to a powerful model. Most do. It is whether the surrounding software can safely provide the context, actions, and feedback that make intelligence useful after the pilot ends.
