F5 says AI inference now core to business operations

F5 has released its latest State of Application Strategy report, which found that 78 per cent of organisations globally now run AI inference themselves.

In Australia and New Zealand, the figure is 48 per cent, pointing to a sizeable gap with the wider market as companies move artificial intelligence workloads into day-to-day operations.

Based on responses from enterprise IT and security leaders, the report suggests AI has moved beyond pilot projects into routine production use. Across the global sample, organisations now manage an average of seven AI models in production, while 77 per cent said inference has become their main AI activity, ahead of model building and training.

By contrast, organisations in Australia and New Zealand reported an average of about four AI models in production. Even so, 72 per cent said they already use AI for automated operational decision-making.

The findings also show how closely AI deployment is tied to broader changes in corporate infrastructure. Globally, 93 per cent of enterprises said they use multicloud setups, while 86 per cent run applications across on-premises, public cloud and colocation environments.

The same pattern is visible in Australia and New Zealand. Three-quarters of organisations in the region said they use multiple cloud providers, and 73 per cent operate across more than one on-premises data centre.

Operational shift

The report argues that AI inference now sits alongside other business-critical systems and must be managed with the same discipline. This marks a shift from the earlier phase of corporate AI adoption, which focused more heavily on experimentation, proof-of-concept work and model training.

According to the research, only 8 per cent of organisations rely solely on public AI services. Most use a mix of models and environments, increasing the need to direct workloads between systems and set policies around cost, accuracy and availability.

Kunal Anand, Chief Product Officer at F5, said the data showed a clear change in how companies are treating artificial intelligence.

"AI has moved from experimentation to operations. The question now is not whether companies will use AI, but whether they can run it reliably, securely, and at scale," Anand said.

He said the implications extend beyond infrastructure.

"This year's data shows a clear shift: AI inference is becoming core to the business, which means AI delivery is now a traffic management challenge, and AI security is now a governance and control challenge. The companies that understand this shift early will be the ones that move faster and more safely," Anand said.

Security pressures

Security emerged as another major theme in the research. Globally, 88 per cent of organisations said they had experienced AI-related security challenges, while 98 per cent said they were preparing for agentic AI, referring to autonomous systems that need identities, permissions and controls.

In Australia and New Zealand, the main barriers were cost and skills. Some 44 per cent of organisations in the region identified skills gaps and the high cost of AI workloads as their biggest obstacles, yet 97 per cent said they were still preparing for agentic AI.

The report also suggests that control points for AI systems are shifting away from infrastructure alone towards prompts, tokens and application programming interfaces. In Australia and New Zealand, 31 per cent of organisations identified prompt layers as the main delivery mechanism, while 24 per cent said token layers were the priority for delivery and security.

That shift matters because it changes where companies must apply oversight. As AI systems are woven into customer services, internal tools and automated decisions, businesses face greater pressure to manage who can access systems, how requests are processed and what safeguards are in place.

Regional lag

The lower level of self-run inference in Australia and New Zealand may reflect a mix of cost constraints, skills shortages and slower deployment cycles. At the same time, the data indicates local organisations are not standing still, given the high level of preparation for agentic AI and the strong uptake of hybrid and multicloud infrastructure.

The regional figures point to a market engaged in AI adoption, but not yet moving at the same pace as global peers in operating inference directly. For vendors, cloud providers and corporate technology teams, that may sharpen the focus on deployment models that reduce complexity while still meeting governance and security requirements.

For business leaders, the broader message is that AI is increasingly being treated as part of core operations rather than a separate innovation track. As organisations spread applications across public cloud, on-premises systems and colocation sites, AI deployment is becoming more closely tied to network management, policy control and cyber security.

Across the survey results, that combination of AI adoption and hybrid infrastructure appears to be setting the direction of travel. Globally, only 8 per cent of organisations now rely exclusively on public AI services.