The AI Herd Is Heading for a Cliff. Here’s Why Sabio Isn’t Following.
There’s a famous line doing the rounds in AI circles right now.
Yann LeCun — Turing Award winner, former Chief AI Scientist at Meta, and one of the architects of modern deep learning — has warned that the technology industry is experiencing a “herd effect.”
Everyone is charging in the same direction. And he believes that direction leads to a dead end.
Writing in the New York Times recently, LeCun was blunt: “There is this herd effect where everyone in Silicon Valley has to work on the same thing. It does not leave much room for other approaches that may be much more promising in the long term.”
His target? Large language models. The very technology underpinning ChatGPT, the AI boom, and the billions being poured into enterprise transformation programmes right now.
At Sabio, we see that herd effect up close. RFPs that start with “Which model should we use?” instead of “Which problem should we solve?” Pilots that never leave the lab because they were designed around a model, not an outcome.
For those of us working at the sharp end of customer experience and contact centre operations, this isn’t abstract philosophy. It has direct and urgent implications for how organisations approach AI investment — and whether those investments will actually deliver.
Powerful Tools. Not Magic Solutions.
Let’s be clear: LLMs are genuinely impressive. They can summarise, generate, translate, and converse at a level that would have seemed impossible five years ago. But LeCun’s critique cuts to something important — these systems don’t understand the world. They don’t reason causally. They don’t plan.
They pattern-match at extraordinary scale, but they do so without the contextual grounding that real operational environments demand.
In a contact centre, that distinction matters enormously. A customer calling about a disputed transaction, a change to a complex insurance policy, or a service failure with compounding knock-on effects isn’t presenting the AI with a pattern-matching problem. They’re presenting it with a real-world problem that requires judgment, context, and action.
That’s precisely why chasing the biggest, shiniest LLM can be the wrong race to be running.
The most significant evolution we’re witnessing in AI right now isn’t about model size, it’s about capability type.
We’re no longer asking AI what, where, or when. We’re asking it to create, do, and build. We’re asking it to use tools, take context, read files, and act. An AI that could perceive became one that could generate. An AI that could generate became one that could reason. And an AI that can reason is now becoming one that can genuinely do work, breaking down problems agentically, reflecting them, and executing solutions.
This is the transition from AI as a novelty to AI as an operational capability. And it’s where the real value lies for contact centres.
Sabio’s Position: Pragmatic, Outcome-Focused, Proven
At Sabio, we’re not trying to build superintelligence. We’re not in a race with OpenAI or DeepMind. Our focus — and our expertise — is relentlessly practical: how do we help organisations deliver better customer experiences, reduce operational friction, and empower their people to do more meaningful work?
LeCun’s warning actually reinforces a view we’ve held for some time. Value doesn’t come from deploying the latest model. It comes from applying the right mix of AI capabilities, governed properly, and aligned to genuine customer and operational outcomes. LLMs are an important component of that mix, but they are not the whole system.
We see this validated constantly. Many AI initiatives fail not because the technology doesn’t work, but because organisations are solving the wrong thing.
In basic language, they’re chasing the engine when they should be designing the car.
The orchestration, the governance, the integration, the human handoff — that’s where transformation actually happens.
The Agentic Opportunity
The most forward-thinking organisations we work with are already shifting their attention from models to agents. They’re asking not “which LLM should we use?” but “how do we build AI systems that can take context from our CRM, reason about a customer’s history, decide on the best course of action, and execute it — with the right guardrails in place?”
That’s agentic AI. And it represents a fundamental shift in what’s possible within the contact centre — not as a future aspiration, but as a deployable reality today.
Think of an AI‑driven system that can triage inbound interactions, retrieve and summarise previous conversations, propose the next best action, trigger workflows in downstream systems, and loop in a human agent when required — all within clear governance and guardrails. That’s a very different proposition to “let’s bolt a chatbot onto our website and hope containment goes up.”
LeCun may be controversial in the Silicon Valley consensus he’s challenging. But from where we stand, working with organisations navigating real operational complexity, his central argument is hard to dismiss: the path to genuinely intelligent, genuinely useful AI runs through understanding, planning, and action, not just prediction.
Let the herd head for the cliff.
We’d rather help our clients build something that actually works. Contact us today for a conversation.