This year marks the 150th anniversary of the telephone. Alexander Graham Bell patented it in 1876, and it was, as far as I’m concerned, the first true man-machine interface — the first time a human being could speak, in real time, through a machine, to another human being who wasn’t standing in front of them.
I’ve spent 48 years in this industry, dating back to working a service desk as a young airman at Elmendorf Air Force Base. I’ve watched voice technology go from electromechanical switching to microsecond digital switching, and soon enough, to whatever comes after that. Through every one of those transitions, one thing stayed constant: the machine’s job was to carry a human conversation.
The detour
Then, sometime in the late 1980s and into the ’90s, we took a detour.
A generation of IT engineers — many of them brilliant, and I count myself among people who understand exactly why this happened — didn’t love talking on the phone. Text felt more comfortable. More controllable. So the industry drifted toward typing as the default way humans and machines, and eventually humans and humans, communicated. The problem is that written language was never built for that. Writing was invented for documentation — record-keeping, history, contracts. It was never designed to carry a live, two-way conversation the way voice does. We took a tool built for one job and asked it to do another, and we’ve been quietly paying for that mismatch for thirty years.
Voice comes back
Now, 150 years after Bell’s patent, AI’s natural language models are bringing voice back around. You can talk to a machine again — not type at it, talk to it — and have it understand you, respond, and help you get something done. We’ve circled back to where this whole industry started, just with a lot more horsepower under the hood.
I didn’t figure this out in a lab. About a year ago, Luiz Domingos, CTO and Group VP Large Enterprise and Vertical Solutions R&D at Mitel, gave me one piece of advice that changed how I work:
Just learn how to talk to it.
Not study it. Not fear it. Talk to it, the way you’d talk to a new technician on your team. I’ve applied that advice every day since — in live service delivery, onboarding, migrations, upgrades, and troubleshooting, with real customers, not in a sandbox. That’s the piece I keep coming back to when people ask me whether AI is actually ready for this industry: it depends entirely on how you’re using it.
Why it isn’t ready to replace support
Here’s the part nobody wants to say out loud: agentic AI is not ready to replace tier-one and tier-two support. Not yet. And it’s not because the models aren’t smart. It’s two specific, fixable problems.
First, hallucinations. Ask the same question to three different AI models and you’ll sometimes get three different answers, because each one is drawing from a different repository of information and weighting it differently.
Second — and this is the one that worries me more — AI is very good at sounding confident even when it’s wrong. If the person asking the question doesn’t know enough to catch a bad answer, they’ll walk away treating a hallucination as gospel truth simply because “the AI said so.” Garbage in, garbage out has never gone away. AI just made it faster and more convincing.
Verify everything
That’s why I don’t think the fix is more automation. I think the fix is alignment — and fundamentals. You still need people who understand what right looks like, who can verify an answer instead of blindly trusting it. I extend the oldest rule I know — zero trust, zero assumptions, verify everything — to AI-generated answers now, the same as I’ve always applied it to people and systems.
That discipline matters more, not less, in an industry that still worships the firefighter. I watched this play out again recently: a technician made a change overnight with no real plan and no rollback strategy, chasing the hero moment, and ended up locking an entire office out of their own systems. That’s not heroism. That’s the absence of the boring, invisible planning that prevents the fire from starting in the first place — the same principle whether the tool in your hand is a punch-down tool or a large language model.
150 years ago, the telephone taught us that the most powerful interface between a human and a machine is the human voice. We forgot that for a while. AI is reminding us. The technology has changed almost beyond recognition since 1876 — but the job hasn’t. It was never really about the wire, or the switch, or the model. It was always about the conversation.
