The Spreadsheet Decade

(Was this newsletter forwarded to you? Sign up here.)

Here's the number worth sitting with if you're worried about what AI does to a marketing team. There are roughly 400,000 fewer bookkeeping and accounting clerks today than there were in 1980. There are also about 600,000 more accountants. The same technology did both (Tim Harford, on BLS data).

That technology was the spreadsheet. VisiCalc shipped for the Apple II in 1979 and was supposed to end accounting. It did the opposite.

Sundar Pichai reached for the same example on The Verge's Decoder in May, asked how close AI was to doing his own job. "I wasn't there when spreadsheets rolled out to companies," he said. "I have to think back to how did people do all this financial analysis before? And I'm sure it changed over a period of three to four years fundamentally, and we got used to it." His read was that the technology "elevates everything to a different foundation."

My read

The spreadsheet didn't shrink finance. It exploded it.

What disappeared was the tabulation: the manual recalculation, the ledger arithmetic, the people whose entire value was getting the numbers to add up. What grew was everything that became possible once a person could model a hundred scenarios before lunch. Leveraged buyouts. Real-time risk. The whole modern apparatus of corporate finance. The category of accountants and auditors hit roughly 1.4 million by 2022. The job that looked threatened was the one that multiplied.

Now look at your own function. A large share of what a marketing team does is tabulation. Building the report. Trafficking the campaign. Cutting the fortieth variant. Reconciling the platforms. Writing the first draft nobody ships as written. That's the layer AI is automating right now, fast.

If your value is the production, the production is what's getting cheap. But the same move finance made is sitting right there. When the tabulation automates, the value migrates up to the judgment. What to make. For whom. Why now. Which idea is worth a quarter of the budget. A model doesn't care whether the work is any good. You do. That's the higher foundation.

The honest part

This is where the optimists get cute, so I won't. The spreadsheet decade was good for finance and bad for the people in finance who only did arithmetic. Both were true at once. The profession got bigger and more valuable, and a tier of jobs inside it did not survive.

Marketing is walking into the same trade. The teams that use AI to do the tabulation faster will compete on cost and lose. The teams that use it to do work that was never possible before, the personalization too granular to staff, the analysis too deep to afford, the campaign too complex to run by hand, will build a kind of marketing the old headcount math never allowed. It's the same skill that wins when building is free: judgment, not production. (More on that in Distribution Is the Last Moat.)

The objection worth taking seriously

Here's the strongest counter. The spreadsheet automated the arithmetic, but a human still did the judgment. AI is coming for the judgment too. So the analogy breaks.

Maybe. Nobody honest knows how far up the stack this goes. But notice that even in the version where AI gets good at the judgment, the scarce thing isn't the analysis, it's the accountability for it: deciding what to bet on and owning the outcome. The financial-analyst job didn't survive because spreadsheets were bad at math. It survived because someone had to choose which model mattered and answer for it. That part doesn't automate cleanly, and it's the part of your job worth protecting.

What I'd do with it

Take your team's actual work and sort it into two columns: tabulation and modeling. Be honest about the ratio. Then ask the only question that matters for the next two years: how fast can you move people from the first column to the second, before the first column automates out from under them?

The accountants who became analysts had a decade to make that move. You have less.