When managers learned to type.

In the 1980s, a fifth of the American workforce handled the clerical and secretarial work. A fifth of the complete working population in America, everyone with a job. Through most of the twentieth century, a large floor of every major company, sometimes a dedicated floor, was given to the typing pool: rows and columns of typists, operating typewriters, and producing the documents their managers could not produce themselves. These typists were trained for this job. Gibbs College, opened in 1911, had been producing great talent at typing. A good typist typed around eighty words per minute, way past what their manager could manage. To reproduce, or make copies of the work, there was carbon paper inserted between the sheets.

But then the word processor arrived, in the late 1970s and accelerating through the 1980s. Offices had started to run with no secretaries, and no typewriters either, replaced by word processors.

Within a decade, the typing pool was effectively gone.
Managers learned the keyboard.

A century-old job category, vanished, not because the work had become unnecessary but because the tool had become operable by the layer above.

The National Secretaries Association feared the new machines would create what they called space-age typing pools. With time it became clear there was no pool at all. The managers simply typed now.

This, I believe, is the state of the modern marketing organisation.

Some years ago, in the early days of performance marketing, our partner agency pitched me multi-touch attribution, a data answer to a marketing question. The question: which channel gets the conversion credit. I was pursuing my postgraduate in AI and ML around the same time. My eyes lit up, I said yes.

We built the architecture to measure every click we could, fed it through a model. The result came back inside a deck presented in its overcompensating gloss and stories. Buried in the gloss was one finding. 60% of what we had been calling organic conversions had actually started on a paid channel. The number was huge, it justified every minute spent on the attribution analysis. And more than that, it gave worth to the time and money spent on my AI/ML program.

Some years later, this marketing x analytics project became my official role. Two reporting lines, one foot in data and one in marketing, pulling specialists from each into one system for customer propensity, intent, and calibrated communication. Something that should not have been the case in a pre-digital era.

It was the tools. Each set of these functions had grown dense enough to fill a career. Illustrator or SQL. Brand or analytics. Holding both became important. So a pre-digital era orthodox marketing split into performance, lifecycle, product, digital, and so on. Data split into engineering, analytics, science, ML for the same reason. Nobody woke up one day and decided marketing should become six disciplines. The tools got hard, and the org chart changed along.

Not just digital marketing. Legal grew a contract-review specialism the moment document software required certification. Accounting split not because the thinking diverged but because the modelling tools did. Engineering fractured into frontend, backend, DevOps, platform the same way marketing fractured. In every field where software did the heavy lifting, the team shape followed the software shape.

This is the part the typing pool explains. The specialisms never emerged because the thinking was different. They emerged because the tools were different. What happens to the thinking layers now that the managers are going to learn to type this time?

What happens when AI collapses
the specialisation needed for tools?

One person working with AI can now run the complete attribution workflow I ran some years ago. The seat that used to direct the data engineer, the analyst and the channel manager can now direct the model, and the model holds the tools. The work has moved up one floor, to the person who was already deciding what the tool should build.

The typing pool is the precedent worth looking at closely here. It happened slowly. A typist who started at Gibbs College in 1965, expecting a forty-year career, could find by 1985 that the entire role had compressed into a feature of the software their manager was now using directly.

This will not be true everywhere. Some specialisms stay because the depth genuinely cannot be held part-time and definitely cannot be managed by hallucinating bots; the cost of being slightly wrong is severe. There the influence stays, and should. But those are the edges of the organisation chart. Most of an org chart is not edges, and the middle is precisely what the typing pool was. The typing pool never held the decision, it held the execution.

The seat deciding what mattered was the decision. The attribution model, or the hybrid analytics and marketing roles, happened with the decisions. I was not in the typing pool then.

So the question is not which roles AI replaces. It is which of your typing pools actually hold the depth, and which are a tool the floor above simply hasn’t learnt yet.

Is genuine depth in the role your truth,
or is it just a tool overseeing the typing pool?