NETWORK PERSPECTIVE
5G promised us flying cars, but what we got was better backhaul
AI faces the same challenge. The models themselves are extraordinary. Nobody doubts the leap forward in reasoning and autonomy. But the hard part isn’t the algorithms, it’s clearing the invisible bars that sit beneath them. Telecoms learned long ago that service levels and accountability can’t be left to chance; AI will need its own equivalent of SLAs before it can scale. Alongside the governance and process, there is also the physical needs of a scaling AI ecosystem. The most valuable companies in the world can build as many data centres as they like, but if the rest of the infrastructure isn’t there, nothing works for the average consumer. 5G promised us flying cars, but what we got was better backhaul. It was a precondition
“who is responsible when a driverless car injures someone in a crash?” It was about governance, accountability and data quality. The physics of going from prompts to autonomous actions adds layers of risk, governance and cost. Layers that standard LLMs don’t need to worry about, but layers that agentic AI will never navigate safely without proper oversight. And this brings us back to why this article appears in the SCTE’s Broadband Journal and not the latest SaaS publication. If historic telecom innovation has taught us anything, it’s that these hidden layers are where real failure hides. Because of these layers, the bar to entering the telecoms arena is as high as Duplantis’s new pole vault world record. And agentic
to the shiny product that was casually overlooked. Agentic AI could be the same: sold as a wild night of disruption, but in practice it may end up folding the laundry. Sexy in the pitch, mundane in the deployment. Yet it’s the everyday graft that separates the one-night stands from the long-term partners. In telecoms, plumbing means ducts, fibre routes and switches. The infrastructure no one sees but everyone depends on. AI needs its own plumbing in governance, data pipelines, integrations and monitoring. Without them, even the smartest agents will collapse under their own weight. Just as Victorian London only thrived once it fixed its sewers, the AI revolution will live or die on what happens below the surface. The telecoms lesson is clear: scale alone isn’t enough. For those who read my previous article, take TalkTalk’s recent breakup as a case in point. The consumer brand, burdened by slim margins and a rapidly changing market, struggled to stay relevant while its wholesale arm thrived because it sat on valuable, investable infrastructure. Those who poured capital into the hidden layers of the network created lasting value and those who chased scale without shoring up the foundations found themselves vulnerable. The same principle applies to AI: success will follow the builders, not the flashy front-ends. Victorian London only took sewage seriously after the Great Stink. AI governance might need its own ‘stink moment’ before we get serious about the plumbing.
Volume 47 No.4 DECEMBER 2025
9
Made with FlippingBook - Online magazine maker