Adapt or Rethink? A Decision Framework for AI Network and Infrastructure Leaders

#networking #architecture #communication #ai# #digital-transformation

The central question today, whether to keep adapting existing networks for AI or rethink the architectures, protocols, and infrastructure underneath them — is rarely answered as a single yes/no choice in the real world. In practice, it is a portfolio decision made based on workload, budget cycle by budget cycle, and it is frequently not optimal: the solution is either by default to incremental upgrades because they are familiar and low-risk, or chase clean-slate rebuilds without a credible transition path.

This talk presents a practitioner-oriented decision framework for making that call deliberately. Drawing on process and digital transformation methodology rather than just protocol design, it maps workloads against two axes — AI-intensity of the traffic pattern and the organization's growth trajectory — to produce four distinct postures: adapt and optimize, adapt now while planning to rethink, selectively rethink critical paths, and rethink the foundation outright. Each posture carries different implications for cost, risk, talent, and speed to value, which the talk lays out side by side.

While the presentation covers examples that can apply across industries, two illustrative examples ground the framework in practice include - In healthcare, a regional hospital network deploying real-time diagnostic AI — imaging triage, sepsis prediction — across facilities sits in the "selective rethink" quadrant: patient-safety latency requirements justify a purpose-built, low-latency fabric for the diagnostic pathway itself, while the broader hospital IT network continues to adapt incrementally, with resilience and forensic auditability of AI-assisted decisions treated as first-class design requirements rather than afterthoughts. In education, a university research computing center scaling distributed AI training across departments and connecting to national research and education networks sits in the "rethink the foundation" quadrant: sustained growth in AI-driven research and curricula justifies an AI-native fabric purpose-built for collective communication, decoupled from — but still interoperable with — the general campus network.

The session closes with a phased roadmap (instrument and baseline → pilot AI-optimized fabrics on critical workloads → architect purpose-built domains) intended to help infrastructure leaders sequence investment without freezing the organization mid-transition — directly addressing the workshop's stated goal of identifying what should be retained, what should be redesigned, and what should be invented.