F5 Report Shows AI Inference Has Become a Core Enterprise Workload

Enterprise AI is moving out of the lab and into daily operations, according to F5’s 2026 State of Application Strategy Report.

The report found that 78% of organizations now run AI inference themselves, showing that enterprises are increasingly choosing direct control over AI workloads rather than relying only on public AI services. F5 said the shift makes AI inference a production workload that now requires the same reliability, security, and governance expected from other mission-critical systems.

For technology leaders, this is an important signal. AI adoption is no longer just about experimenting with models or testing proofs of concept. The bigger challenge is now how enterprises deliver, secure, route, monitor, and govern AI across complex infrastructure.

AI Is Becoming Part of the Application Stack

F5’s research shows that organizations are now coordinating an average of seven AI models in production. The report also found that 77% of organizations say inference has become their dominant AI activity, surpassing model building and training.

That changes how enterprises need to think about AI operations.

Inference is the process of running trained models to generate outputs. Once inference becomes part of production, it must be treated like any other application workload. It needs uptime, performance management, access controls, routing logic, cost controls, observability, and security policies.

Kunal Anand, Chief Product Officer at F5, said the question is no longer whether companies will use AI, but whether they can run it reliably, securely, and at scale. He also said AI delivery is becoming a traffic management challenge, while AI security is becoming a governance and control challenge.

Public AI Services Are Not the Only Path

One of the most notable findings is that only 8% of organizations rely exclusively on public AI services. Most enterprises are building more diversified AI model portfolios.

That matters because different models can serve different use cases, cost profiles, latency needs, accuracy requirements, and security constraints. As enterprises use multiple models, they need better routing, fallback, and policy controls.

This makes AI infrastructure more similar to modern application delivery. Enterprises need to decide where workloads run, which model handles a request, how to manage failures, and how to enforce policies across environments.

Hybrid Multicloud Is Now the Default Environment

The report also shows how AI is being deployed into already complex IT environments.

According to F5, 93% of organizations use multiple clouds, while 86% run applications across hybrid multicloud environments, including on-premises, public cloud, and colocation infrastructure.

This creates a major operational challenge. AI workloads must work across distributed environments without creating new silos. Enterprises need consistent delivery, security, identity, and governance controls across every place where applications and AI systems run.

For CIOs, CISOs, and platform teams, this means AI infrastructure cannot be managed separately from application infrastructure. AI delivery and application delivery are becoming part of the same operating problem.

AI Security Is Becoming a Systemic Requirement

F5’s report also highlights the security pressure created by production AI.

The report found that 88% of organizations have faced AI-related security challenges. It also found that 98% are preparing for agentic AI, where autonomous systems may need identities, permissions, and guardrails similar to human users.

This is a major shift for security teams. Traditional application security models are not enough when AI systems interact through prompts, APIs, tokens, identities, and autonomous workflows.

As agentic AI becomes more common, enterprises will need to govern what AI agents can access, what actions they can take, how they authenticate, and how their decisions are monitored. Without that control layer, AI adoption can increase operational and security risk.

Prompts and Tokens Are Becoming Control Points

F5’s report also points to a new area of enterprise control: the prompt and token layers.

Nearly 29% of organizations identified prompt layers as the top delivery mechanism, while 23% prioritized token layers for delivery and security.

This shows that AI delivery is not only about servers, networks, or APIs. Enterprises also need to manage the actual interaction layers that drive model behavior and cost.

Prompts influence what AI systems produce. Tokens affect cost, performance, throughput, and security. As AI usage grows, controlling these layers will become important for governance, compliance, safety, and budget management.

Why This Matters for Enterprise Technology Leaders

The F5 report shows that enterprise AI maturity is becoming tied to operational resilience.

Companies are no longer asking only how to use AI. They are asking how to run AI in production across multicloud systems, secure it across multiple control points, and govern it as part of the application stack.

For B2B technology buyers, the practical takeaway is clear: AI infrastructure decisions now need to include application delivery, security, observability, identity, routing, and policy management.

The companies that manage this well may be able to move faster with AI while reducing risk. The companies that treat AI as a standalone experiment may struggle as inference workloads become more central to business operations.

F5’s report suggests that the next stage of enterprise AI will not be defined only by model capability. It will be defined by how reliably, securely, and efficiently organizations can operate AI across real production environments.

Key Source / Reference

Official source: Business Wire — AI Has Left the Lab: F5 Report Reveals 78% of Enterprises Now Run AI Inference as a Core Operation

FAQ Section

What did F5 announce?

F5 released its 2026 State of Application Strategy Report, which shows that AI inference has become a core enterprise operation rather than only an experimental activity.

What percentage of enterprises now run AI inference themselves?

According to F5, 78% of organizations now run AI inference themselves.

Why is AI inference important for enterprises?

AI inference is important because it is where trained models generate outputs for real business use. Once inference becomes a production workload, it needs reliability, security, routing, observability, and governance.

How many AI models are enterprises running in production?

F5’s report says organizations are coordinating an average of seven AI models in production.

What does the report say about hybrid multicloud?

The report says 93% of organizations use multiple clouds and 86% run applications across hybrid multicloud environments.

Why does agentic AI create new security requirements?

Agentic AI systems may act more autonomously, which means they need identities, permissions, guardrails, and governance controls similar to human users or service accounts.