Only 28 Percent of Companies Manage AI From a Single Point, F5 Report Finds

Companies around the world are rolling out artificial intelligence into their daily operations at a rapid pace, but only a handful are doing it in an organized way. F5's latest report finds that just 28 percent of organizations manage their AI deployments from a single, centralized control point. The rest are building an increasingly fragmented technology landscape that is growing harder to oversee.

What the Report Found

F5, a company specializing in application security and delivery, surveyed hundreds of IT and security leaders at enterprises around the world. The conclusions are clear: artificial intelligence has stopped being an experiment and become part of everyday business infrastructure, but management practices haven't kept pace with the speed of deployment. A full 78 percent of organizations now run AI inference as a core operation, and the average company is coordinating seven models running simultaneously in production or in preparation for deployment.

The problem is that these models rarely live in one coherent system. Instead, they end up spread across different teams, clouds and tools that don't always talk to each other. The result is fragmented accountability and no single place to check exactly what a given AI instance is doing, who has access to it, and what data flows through it.

Fragmentation as the Norm

The report's authors stress that the problem isn't a lack of will to bring order to these systems, but the nature of the technology itself. AI models enter organizations through different doors: some arrive via IT department initiatives, some through individual product teams testing new tools, and some through external cloud service providers. Each of these paths generates its own access rules, its own logs and its own security gaps. As a result, 88 percent of surveyed organizations admit that integrating AI outputs into daily operations comes with architectural, organizational or security challenges.

Authentication and Model Protection

Among the specific technical challenges, respondents most often pointed to identity and access issues. 55 percent of companies see managing authentication and controlling access to the APIs through which models communicate with the rest of the infrastructure as a priority. These interfaces often become the weakest link, since they tend to be added quickly, under time pressure, without a full security review.

Another 54 percent of companies openly admit they lack sufficient visibility into what AI resources are even being used across the organization. Half of respondents, 51 percent, point to efforts to prevent information leaks, while 43 percent focus on protecting the models themselves from misuse, such as manipulation of the prompts sent to the system.

AI doesn't remove existing organizational complexity. If anything, it can amplify it - Bartłomiej Anszperger, CTO at F5

F5's Global Perspective

In commentary accompanying the global version of the report, F5 Chief Product Officer Kunal Anand describes the shift in how companies have approached AI over the past year. According to him, artificial intelligence has moved past the experimentation stage and entered an operational phase, meaning it now needs to be treated with the same rigor as business-critical systems. The report also highlights the scale of the environments enterprises operate in today: 93 percent of organizations work in hybrid and multi-cloud environments, and 98 percent are already preparing to deploy agentic AI, meaning systems capable of independently carrying out multi-step tasks.

AI has moved from experimentation to operations. AI inference is becoming a core part of the business, which means delivering AI is now a traffic management challenge - Kunal Anand, Chief Product Officer at F5

What It Means for Companies in Poland

For Polish businesses, the report's findings carry practical weight, since a similar pattern of chaotic AI adoption is visible locally too. Companies are testing chatbots, assistants and analytical models within individual departments, often without involving security teams, which leaves no one in the organization with a full picture of which AI systems are actually running and what data they process.

Growing regulatory pressure, including the EU's AI Act, further raises the cost of this fragmentation. Companies that cannot pinpoint where and how they use AI will find it harder to demonstrate compliance with rules requiring audits and documentation of high-risk systems. Consolidating management, as the F5 report puts it, is therefore becoming not just a matter of efficiency but also of preparing for regulatory scrutiny.

As a way out of this impasse, the report's authors point to building a unified control point covering identity management, API access, monitoring of token-based costs, and protection against prompt manipulation targeting the models. Without such an approach, the fragmentation of AI systems will keep growing in proportion to the number of models deployed, along with the associated costs and risk of security incidents.