In this report, we discuss the extent to which AI adoption will transform enterprise networking and set out a framework of key questions that decision-makers should consider when developing and designing their AI strategies.

Widespread adoption of four main types of AI applications will have a significant impact on enterprise networks

In this research, we have broken down AI applications into four main types that will affect enterprise networks in various ways.

1. Off-the-shelf, LLM-based productivity tools which can but don’t have to be trained and fine-tuned using proprietary enterprise data (e.g., ChatGPT, Microsoft Copilot, DeepSeek). Most processing happens in the cloud (private/public), so impact on enterprise WAN is low. Some use cases, however, will require huge data volume upload (e.g., code reviews) that could strain networks. Inferencing using proprietary data also increases security requirements. Ensure the network is a facilitator rather than a roadblock.
   
2. Existing software applications now infused with AI to enhance key functionalities and capabilities (e.g., Salesforce’s Agentforce). These applications are typically cloud-hosted, so impact on enterprise WAN is limited. However, because AI features may be adopted without formal approval processes, their network impact can be unaccounted for posing a hidden risk or ‘silent killer’. Proactively check with IT and business teams to monitor AI adoption and its impact on network performance.
   
3. Additional dedicated AI solutions that are not specific to a given vertical (e.g., customer chatbots, hyper-personalised marketing, cybersecurity). These solutions are often vetted before deployment, allowing teams to plan for network impacts. Teams should monitor data flows, especially if the applications run at the edge. Assess the expected impact of application on network based on its specific data requirements, especially if processed at the edge.
   
4. Dedicated AI solutions specifically designed for a given industry vertical (e.g., computer vision for theft prevention in retail). This category includes a wide variety of use cases, so impact varies. Video-heavy or ‘data-unlocking’ applications may drive large increases in traffic volume and significantly shift traffic patterns, especially where AI enables enterprises to process previously unused or discarded data, such as archived video or sensor logs. Be aware of your enterprise AI adoption roadmap and proactively assess capacity.

If you are not a subscriber, enter your details below to download a free copy of the report

Furthermore, agentic AI systems that require on-demand access to data across diverse infrastructure, such as on-premises servers and edge devices, can significantly increase network demands. In multi-agent environments, low latency is critical for smooth coordination, while high reliability is essential for mission-critical applications that cannot tolerate downtime. Remain alert: agentic AI is fundamentally different from traditional enterprise systems, as it acts autonomously and can drive dynamic, unpredictable traffic patterns.

The location of AI workload processing is largely dictated by the type of enterprise AI application

Table of contents

• Executive summary
• Impact of location of AI workload processing on enterprise networks
• Four main types of enterprise AI applications
• Off-the-shelf LLM-based productivity tools that can be trained on proprietary enterprise data
• Existing software applications infused with AI to enhance key functionalities
• Additional dedicated AI solutions that are not specific to a given industry vertical
• Dedicated AI solutions specifically designed for a given industry vertical
• Implications of agentic AI applications
• Conclusion and recommendations
• Message from our research sponsors

Related research

• Connected health: Enabling smarter, secure and scalable care
• Simplifying NaaS for easier enterprise consumption: Capturing the USD100bn opportunity
• Real-time automation: The future of transport & logistics
• SMBs: How telcos should approach them

If you are not a subscriber, enter your details below to download a free copy of the report

Harine Tharmarajah

Harine Tharmarajah

Strategy Consultant

Harine is a Consultant at STL Partners, who joined after completing her undergraduate studies. She earned a First class honours degree in BSc Economics from University College London. Since joining STL Partners, Harine has worked with telecoms and technology companies on strategic engagements in network transformation, but also works within STL's Edge practice.