What Happened to Doris on the Switchboard?

AI probably won’t replace work in one dramatic moment. More likely, it will do what technology has always done: quietly remove pieces of work until one day we look around and realise the job has changed.

When I started my career, every office had a Doris.

She was not always called Doris, of course. Sometimes she was Margaret. Sometimes she was Sheila. But if you worked in an office before everything became an app, you know exactly who I mean.

Doris sat somewhere near reception. She knew every extension number in the building. She knew who was on holiday, who was in a meeting, who had popped out for lunch, and who absolutely did not want to take that particular call. More impressively, she often seemed to know who was calling before they had finished introducing themselves.

Doris was an intelligent routing engine.

She just happened to be human.

Today, Doris has gone. In fact, if you are under thirty, there is a decent chance you have never met a Doris. You may never have seen a switchboard. You may never have used a desk phone in any meaningful way. Some younger people reading this are probably wondering why any organisation ever needed a dedicated person just to answer telephones.

Which is exactly the point.

Nobody woke up one morning and announced a strategic initiative to replace Doris. Nobody declared war on switchboard operators. Nobody sat in a boardroom and said, “Let’s get rid of reception.” What happened was quieter than that. Technology crept forward.

The switchboard became a corporate directory. The directory became a digital phone system. The digital phone system became VoIP. The desk phone became a softphone. The softphone became an app. Then one day we all looked around and asked, “What happened to my phone?”

The answer was the same as it usually is with technology: hundreds of small improvements, each one sensible, each one logical, each one saving a little bit of time. Eventually, the world changed.

That is why I sometimes smile when people argue about artificial intelligence.

One side insists AI will take everyone’s job. The other insists it will not change anything important. History suggests both camps are probably wrong. Technology rarely arrives carrying redundancy notices. It usually arrives carrying productivity gains.

The redundancy notices turn up years later.

A strange thing happened to me recently. I received an email so obviously written by AI that I could not quite bring myself to finish reading it.

The worst part was that this was not some corporate update from a software vendor or a three-page policy announcement from HR. It was from my son’s A-level teacher.

You know the sort. Three pages when three paragraphs would do. Random words in bold. Sentences that sounded polished but somehow said very little. The unmistakable feeling that someone had asked ChatGPT to “make this sound more professional” and then accepted the first answer.

The irony was that a phone call would have been quicker. In fact, a phone call would probably have taken less time than writing the prompt, reading the output, formatting the response, and sending the email.

Halfway through, I gave up. I pasted it into Claude and asked a simple question: “What is this person actually trying to tell me?”

Within seconds, I had an answer.

One AI had written the email. Another AI had translated it back into English. I sat back and realised we had crossed a line somewhere. The machines are now talking to each other, and we are standing in the middle pretending this is normal.

I call it ChatGPT Tennis with VAR.

At the same time, the technology industry has developed a remarkable ability to invent new words faster than businesses can understand them.

Agentic. Sovereign. Reasoning. Autonomous. Orchestration.

Spend ten minutes on LinkedIn and you would be forgiven for thinking every company in Britain is building an agentic sovereign reasoning platform powered by autonomous agents. I have worked in technology for nearly thirty years, and I am increasingly convinced that half the people using these terms could not confidently explain them after two pints.

That is not a criticism of the technology itself. Many of these concepts are genuinely important. Sovereignty matters. Governance matters. Understanding where your data lives and who has access to it matters. My concern is that we have become better at discussing the terminology than explaining the business outcome.

Every generation of technology creates its own vocabulary. Mainframe. Client-server. Cloud. Digital. Big Data. AI. We spend two years arguing about the definition and ten years quietly deploying it.

My favourite term at the moment is “agentic”. Ask ten people what it means and you will get eleven answers. The simplest explanation I’ve found is that software observes something, makes a decision, and performs an action.

Which immediately raises another question: has software not been doing that for years?

My thermostat observes the temperature, decides whether the heating should come on, and takes action. Nobody called that agentic. They called it central heating.

The problem with all this terminology is that it distracts us from what is actually happening. Most organisations do not have an intelligence problem. They have an attention problem.

Every organisation I visit is drowning in data. Financial reports. Customer records. Energy consumption. Phone statistics. Building information. Operational metrics. The challenge is not collecting information. The challenge is finding the time to notice what the information is trying to tell us.

One of the more interesting examples we have worked on recently involved something deeply unexciting: missed phone calls.

Not artificial general intelligence. Not robots. Not the end of humanity. Missed phone calls.

A Teams phone system was already recording every call, every queue, and every abandoned call. The information had existed for years. Nobody had done anything particularly useful with it because everybody was busy. The data was there, but the attention was not.

The system spotted a pattern. Certain periods of the day consistently suffered from missed calls. Other periods had too many people sitting in call queues when they could have been getting on with other work. The system did not invent anything. It did not replace anybody. It simply noticed.

From there, it could recommend adjustments to queue membership during busy periods and reduce interruptions during quieter ones. More calls answered. Fewer unnecessary interruptions. Better service. The information had been sitting there all along. The organisation simply had not had time to look.

Perhaps that is where I differ from both the AI evangelists and the AI sceptics.

I absolutely think AI will replace jobs, just as spreadsheets replaced armies of bookkeepers, just as online booking systems replaced travel desks, and just as digital telephony eventually replaced Doris. But I suspect it will happen in two very different ways.

The first is obvious. Highly paid professions that operate within clear rules and boundaries will be affected. Parts of law. Parts of pharmacy. Parts of compliance. Parts of accountancy. When work is expensive, repetitive, document-heavy, and governed by rules, the economics are difficult to ignore.

The second form of replacement will be far more subtle, and probably far more common.

A report gets automated. A workflow gets streamlined. A dashboard becomes self-service. A queue becomes self-optimising. An email becomes a process. Each improvement saves a fraction of a percent. Half a percent here. One percent there. Nothing dramatic enough to call a transformation.

Then somebody retires, and nobody recruits a replacement.

Not because AI took their job in the theatrical sense. Not because a robot marched in and sat at their desk. But because the organisation quietly stopped needing quite as much of the work.

That is how technology usually changes employment. Not through one dramatic event, but through accumulation. Tasks disappear first. Then processes change. Then teams reshape. Then roles that once felt permanent start to look strangely historical.

Which brings me back to Doris.

Doris did not lose her job to AI. Doris lost her job to thousands of small improvements spread across thirty years. No single change made her obsolete. Each one simply made the old way of working slightly less necessary.

Perhaps the real question is not whether AI will replace people. Of course it will replace some tasks. Technology has always done that.

The more interesting question is this: what is today’s switchboard?

Which part of your organisation feels permanent, essential, and impossible to automate? Which process is so familiar that nobody questions it? Which role exists because, for years, it has simply been the way things are done?

Because somewhere, right now, somebody is performing a task that feels every bit as important as operating a switchboard once did. And twenty years from now, a new generation will look back and ask:

“What happened to Doris?”

The answer will be the same as it has always been.

Thousands of small decisions. Thousands of sensible improvements. And a future that arrived so gradually that hardly anybody noticed it happening.