Cisco has urged organisations to stop layering AI onto legacy systems, revealing that just 33% of companies have a formal adoption strategy. Drawing on its 2025 AI Readiness Index, the company outlined three core principles for success. First, firms must establish secure data context across fragmented systems. Second, organisations should govern shadow AI by providing secure, model-agnostic platforms. Third, Cisco advocates for complete workflow redesign rather than minor automation. While 83% of firms plan to deploy AI agents, only 31% currently feel equipped to secure them.
Cisco has published guidance on operationalising AI in large organisations, arguing that many enterprises are still attempting to layer AI onto systems and processes that were never designed to carry it.
The company points to its 2025 AI Readiness Index, which found that only 33% of organisations have a formal plan to guide employees through AI adoption. The same study, based on responses from 8,039 senior business leaders in 30 markets including Saudi Arabia and the UAE, found that 83% of companies plan to develop or deploy AI agents, while just 31% say they are fully equipped to control and secure agentic AI systems.
Cisco sets out three principles. The first concerns data: business information sits across applications, warehouses, documents and legacy systems, and Cisco argues that no model will produce useful answers without secure access to the right information and the context around it. That means connecting AI to the applications where data already lives and building the semantic understanding needed for it to reason across the business.
The second addresses shadow AI. When generative AI tools first became widely available, employees at most organisations began using consumer products immediately. Cisco’s position is that shadow AI becomes the default in the absence of a secure alternative, and its internal response was to build a model-agnostic AI platform governed under its Responsible AI Principles rather than attempt to block the behaviour. The platform was designed to be secure enough for staff to work with enterprise data, broad enough to give access to different models for different tasks, and extensible enough for teams to build and share prompts, projects, connectors and agents.
The third principle concerns workflow design. Cisco argues that applying AI to each step of an existing ten-step process delivers less than rebuilding the process around AI from the start. More than 21,000 Cisco engineers now use AI coding tools, saving an average of six hours a week; employees across the wider business save an average of five hours.
On agents, Cisco’s stated position is that the objective is the right level of autonomy for a given task rather than maximum autonomy, with trusted data, secure access, enterprise context and governance becoming more important as AI moves from answering questions to completing work.
