Press 1 for English. Press 2 for billing. Press 3 to speak with a representative. Please hold.
It’s 2026, and most companies still serve their customers with technology from the 90s. IVR – Interactive Voice Response – was already a bad idea back then. It was just the only idea.
It isn’t anymore.
Customers have already moved on
The Voices.com Amplified 2026 report surveyed 700 business leaders and consumers about voice AI. The results are fairly clear: 55 percent of consumers already use voice to interact with AI. 65 percent say it’s faster than typing.
The interesting part isn’t the numbers themselves. It’s what’s happening on the other side: only 29 percent of companies have actually deployed voice AI.
There’s a gap. And it’s not a technical gap. The technology is there. Prices have dropped. Quality has reached a point where most people can’t tell the difference between an AI voice and a human. The gap is a leadership decision that hasn’t been made.
The market isn’t waiting
In March, Zendesk acquired Forethought – a company building self-learning AI agents for customer service. Zendesk’s own AI agents already resolve over 80 percent of customer queries automatically. Now they’re adding voice.
ServiceNow launched their Autonomous Workforce – AI specialists handling over 90 percent of IT requests and resolving them 99 percent faster than humans.
These aren’t prototypes. They’re production systems at large enterprises running right now.
And then there are the companies still asking their customers to press 1.
Why it’s taking so long
It’s tempting to say companies are just slow. But that’s too simple.
The problem is that voice AI isn’t an IT project. It’s an organizational project. An AI agent answering customer queries needs access to order data, customer history, return policies, inventory status – and it needs to act on it. Not just read from an FAQ.
That requires systems talking to each other. And as I’ve written about before, that’s exactly where most organizations get stuck. Not because the technology is missing, but because data lives in silos that were never designed to share.
So companies choose the safe option. Another IVR menu. Another chat widget with three pre-programmed answers. Customers sigh and call back Monday.
It’s about who answers the phone
There’s a deeper question here than technology.
When 55 percent of your customers are already used to talking to AI – and experience it as faster and better – it changes their expectations of you. Not in a year. Now.
Every time a customer calls your company and hits an IVR menu, they’re comparing that experience to the one they had with a competitor, a bank, or just when they asked their phone about the weather. And your IVR loses that comparison. Every time.
This isn’t a question of buying the right software. It’s a question of deciding that the way customers experience your company is actually important enough to do something about.
Where to start
Not by buying a platform. Start by mapping what customers actually call about. Most companies have between five and ten enquiry types that make up 80 percent of all calls. Half of those can be handled by an AI agent today – if it has access to the right data.
That’s a bounded project. Not a transformation. A pilot that proves the value and reveals what’s missing in your data architecture.
The alternative is to wait. And while you wait, your customers are learning to talk to AI with everyone else.
