Voice as a digital channel

Walk into almost any contact centre and you'll find the team split in two: voice on one side, digital on the other. Different leaders, different tools, different metrics, often different floors. Customers don't experience their lives that way — they pick up the phone, then they open the app, then they reply to the email, all about the same problem — but the org chart keeps insisting these are separate channels with separate disciplines.

That line was never really about customer behaviour. It was about cost.

Why the line existed

Voice was expensive to experiment with. To try something new on a phone call you needed telephony hardware, an IVR licence, professional services to wire it up, and a long change-control process to put it anywhere near a production queue. A small idea took a quarter to test. The same idea in chat or email took an afternoon.

So digital became where teams iterated, and voice became where teams kept things stable. The split wasn't strategic. It was the natural consequence of one channel being cheap to change and the other being expensive to touch.

What changed

Voice is now cheap to change. Cloud-based contact centre platforms, programmable voice APIs, and — increasingly — language models that can produce and understand speech, have collapsed the experimentation cost. The MVP that someone in your business sketched on a whiteboard last week can be a working voice prototype this week and in front of real callers next week. That wasn't true three years ago.

Two consequences follow, and both matter more than the technology itself.

First, the pace mismatch is gone. When voice and digital iterate at the same speed, there's no longer a structural reason to keep them in separate org silos. The teams can — and should — merge around the customer's intent rather than the channel's plumbing.

Second, AI is now doing some of the testing. One of the genuinely interesting developments isn't that AI handles calls; it's that AI tests the systems that handle calls. Does the call get answered? Is the audio quality acceptable? Was the model's answer accurate, or a hallucination? Are the references it cited still valid? These were manual QA tasks. They're rapidly becoming automated, which means you can deploy voice changes with the kind of confidence that used to be reserved for software releases.

"Voice" isn't going to mean what it used to mean

Here's the part most strategy decks miss. When people talk about voice in CX, they almost always mean the phone call. A customer dials a number; an agent or an IVR picks up.

The interesting growth isn't there. It's in voice and audio embedded inside everything else — a voice mode in your mobile app, audio inside a chat thread, a video call that escalates from a web form. The phone call is one delivery mechanism for a voice interaction. There are about to be many.

If you're still planning your voice strategy as "the phone channel," you're optimising the smallest version of the surface you're going to have to support.

The point of all this

It's tempting to read everything above as a technology argument. It isn't. The technology is the easy part.

The reason any of this matters is what happens in the rare, good moment when a voice channel works well. Someone calls in frustrated — angry, maybe uncivil. On the other end is a person (or a thoughtfully designed assistant) who actually listens, actually understands, and has the right information at hand to solve the problem. The wall between the customer and the company comes down. The agent ends up sitting beside the customer, working on the problem with them rather than across from them.

That moment turns detractors into loyal customers more reliably than any other interaction your company will ever have. It's the highest-leverage thing a contact centre can do.

For decades, the voice-vs-digital split made that moment harder to engineer. The line existed for cost reasons, but the cost it imposed wasn't on the company's budget — it was on the customer's experience. Now that the line is coming down, the question worth asking inside your organisation isn't "how do we add AI to voice?"

It's: "now that voice can change as quickly as everything else, what would we build if we stopped treating it as a separate channel?"

This piece is drawn from a talk I gave on AI and customer experience in 2026. The full talk explores several other threads — customer rage, learned helplessness, AI augmentation of human agents — that I'll be writing up separately.