Peak Contact Centre

There's a particular moment in history when a civilisation reaches its architectural zenith — just before the model that made it possible becomes obsolete.

I think we're living that moment in the contact centre industry right now.

We have never been more sophisticated. Workforce management tools that forecast to the quarter-hour. IVR trees mapped to every conceivable intent. ACD logic that would make a logistics engineer weep with admiration. Skills-based routing tuned across dozens of queues. A sprawling ecosystem of platforms — Genesys, NICE, Cisco, Avaya — each offering more features than any single deployment will ever fully use.

We are at peak contact centre. And I'm genuinely asking: is that a good thing?

The Architecture Was Built Around Scarcity

Everything we designed — IVRs, queues, routing rules, priority tiers, overflow logic — was built to solve one fundamental problem: agents are finite and expensive.

If you can't answer every call immediately, you need to triage. You need to understand intent as cheaply as possible (press 1 for billing). You need to rank customers by value or urgency. You need to predict volumes and staff accordingly. You need to manage the wait so people don't abandon.

That entire operational model is a sophisticated, decades-refined response to the scarcity of human attention.

What happens when attention is no longer scarce?

Unlimited Agents Changes Everything — If You're Willing to Accept It

AI voice and chat agents aren't just faster, cheaper humans. They're a fundamentally different resource model.

Concurrent capacity is effectively unlimited. There is no queue. There is no "all agents are busy." There is no Monday morning spike that requires you to have hired 40 extra staff in October for Christmas cover.

And if you don't have a queue, why do you have a queueing system?

If an AI agent can establish intent through three natural language exchanges in fifteen seconds, what exactly is the IVR for? It was always a blunt instrument — a DTMF-based approximation of understanding, dressed up in professional voice talent. We tolerated it because the alternative (putting every caller straight to a human) was economically impossible.

That constraint is dissolving.

The Rise of LAMs Changes the Question Entirely

We've spent the last two years debating LLMs in the contact centre — which model, which vendor, RAG vs fine-tuning, hallucination risk, compliance implications. Important questions. But they're still framing AI as a better tool inside the existing model.

Large Action Models are a different proposition.

An LLM understands your intent. A LAM does something about it.

When an AI agent can not only comprehend that you want to move your direct debit date, but can authenticate you, query your account, execute the change, confirm it, and log the interaction — all within a single conversation — the contact centre stops being a place where humans resolve things and becomes infrastructure for exceptions the AI hasn't been trained on yet.

That's not an evolution of the contact centre. That's a redefinition of its purpose.

So What Are We Actually Building Toward?

I don't think the contact centre disappears. But I think the shape of it changes dramatically.

The high-volume, repeatable, intent-clear interactions — the 60–70% of contacts that any mature operation has already tried to deflect — those will be handled autonomously and well. Not as a cost-cutting exercise dressed up in customer experience language, but because the experience genuinely improves. No hold music. No "your call is important to us." Just resolution.

What remains is genuinely complex. Emotionally charged. Ambiguous. High-stakes. The interactions where human judgement, empathy, and accountability actually matter — a customer in financial hardship, a bereaved family member sorting out an account, a business owner whose service failure is costing them real money by the hour.

This is not the death of humans in CX. Far from it.

But it is a profound shift in what we ask of humans. The agent of the near future isn't fielding their fiftieth password reset of the day. They're handling the call that actually needed a person — and they're doing it with something today's agents rarely have: complete context. Full interaction history. Sentiment trajectory across every prior touchpoint. Account data, product data, case data — surfaced in real time, not buried across four systems and a sticky note.

An empowered human, handling fewer but more meaningful interactions, armed with data that was previously impossible to assemble in the moment — that's a better job. It's also a better experience for the customer.

The humans don't disappear. They step forward. The noise clears, and what's left is the work that actually requires them.

The Question Nobody Wants to Ask Out Loud

The contact centre industry has genuinely matured. NPS, sentiment scoring, and first contact resolution have largely displaced the blunt AHT obsession of a decade ago. Operators are more sophisticated. The intent is right.

But here's the uncomfortable truth: even the best-tuned FCR metrics and sentiment dashboards are still measuring how well we manage a process that may no longer need to exist in its current form. We've gotten very good at measuring performance within a model designed around scarcity. The metrics improved. The underlying architecture didn't change.

Optimising a queue beautifully is still optimising a queue.

If we designed this from scratch today — with unlimited concurrent AI capacity, natural language intent resolution, LAMs that can execute across live systems, and humans freed up for the interactions that genuinely need them — what would it actually look like?

Almost certainly not this.

That's not a criticism. It's an invitation. The organisations that ask that question now, before they're forced to, will define what comes next.

Paul Wilson — Co-founder, Canzuki | Vendor-agnostic CX consulting across NZ & AU | Problem first. Platform last.