The Automation Accelerant

Policy that pushes up wages and costs speeds automation beyond its natural pace, the pace the technology alone would set. Technology sets whether the job gets automated. Where the machine is mostly hardware and software, its price to the buyer trends down year after year.¹ A vote can slow that locally for a while, but it doesn’t turn it round. What the policy changes is when. Anything that raises the cost of employing a person, wherever a machine can already do part of the job, pulls the automation date toward the present, sometimes by years.

The mechanism first

Nearly every automation decision inside a business is the same piece of arithmetic. One side of the ledger is the all-in cost of a person doing a task. That’s the wage, plus the employer’s National Insurance, pension contributions, cover for holidays and sickness, training, and staff turnover that the employer carries on top of it.² The other side is a machine that can do some of the task. The machine’s side falls year after year because hardware and software get cheaper with scale and competition.³ Where the buyer buys a fitted and serviced package, no published price index covers it, but the goods inside it fall on that trend. The fitting and the support are work done by people on local wages. As the tools improve, that work takes fewer hours. A supply shock [a sudden shortage or jump in costs] can hold the goods part up, or push it back up, for about a year.⁴ Where the government and central bank keep expanding money and credit, the goods part stays up for longer. Underneath, the cost of making those goods keeps falling.

The person’s side of the ledger is where policy acts, through two doors. The direct door is mandated cost, and it arrives as a higher wage floor, an employer’s National Insurance rise, or a new benefit the law requires. The indirect door is the government and central bank expanding money and credit, which pushes up the café’s rent, beans, and energy, and pushes staff to ask for more because their own rent and food cost more. Both doors raise the person’s side. No law puts a wage floor on the machine, so the direct door reaches it only through the pay of the people who make, ship, and fit it. For a kiosk [a self-order screen] that’s the fitting and the support, not the chips or the code, and it’s small beside the person’s side of the ledger, where the whole of the cost moves with the floor. Its running costs, the power and the software fee, can still be pushed up by the same money expansion that lifts the café’s other bills. Its purchase price keeps falling. So the crossover, the day the machine beats the person on cost, arrives earlier than it would have. The automation was always coming. The policy moves the date.

The café

Take a small independent café with three people on the morning shift, one on coffee, one on food, one on the till taking orders and payments. The till job pays somewhere near the wage floor, the least an employer may legally pay, call it £12.70 an hour.⁵ Add the employer’s National Insurance and pension contribution and the true cost is closer to £15 an hour before holiday cover and training. At full-time hours that’s roughly £30,000 a year to have a person taking orders. Rough numbers, but the shape is right.

The alternative is a pair of self-order screens with card payment built in. A few thousand pounds up front and a modest monthly software fee, call it £5,000 in the first year and much less after. The screens work every opening hour and never call in sick. If they let the owner run each shift with one fewer person at the counter, the saving is around £400 a week, over £20,000 a year. That’s the shift’s counter hours going, not a whole post. The package pays for itself in about three months. Once the sums look like that, the owner is choosing between installing the screens and watching the café across the road install them first and serve the same flat white for 20p less.

In the same café ten years back, the wage floor was around £7 and the all-in cost maybe £8.50 an hour.⁶ The screens cost several times more and did less, with clunky software and card payments that didn’t work well. Payback was measured in years, so most owners didn’t bother. Between then and now, both lines moved. Technology cut the machine’s price, which was going to happen anyway. Policy raised the person’s price, with the wage floor lifted by law year after year, the employer’s National Insurance rise in April 2025 on top for every employer bar the very smallest, and the general rise in living costs pushing wage demands up across the board. Every one of those moves dragged the crossover date toward the present. The natural pace, by which I mean the date set by the technology alone under honest prices, might have put screens in that café around the end of the decade. The policy-loaded ledger put them in this year. Same destination, earlier train.

Why the owner can’t dodge it

Some owners raise prices instead, and they lose custom. And the competitor who automates can hold prices down or improve service, which pulls the whole street toward the screens. Margins in hospitality are typically only a few pence in the pound once everything is paid, so a cost rise that can’t be passed on has to come out of something. It comes out of the way the work is done, or out of the quality and the quantity of what’s sold, or out of the owner’s own profit and unpaid hours for as long as those last, or out of what the owner saves when better-paid staff stay longer, or the business closes.⁷ Cruelty doesn’t come into it. The owner is reading a ledger the whole street is reading too.

What it looks like from the worker’s side

Automation rarely looks like a sacking on a Tuesday. It looks like fewer hours on the rota, the person who leaves not being replaced, one person running a counter that used to take two, and the next branch opening with a leaner staffing plan from day one. Owners tend to make the switch when a cost shock or a downturn forces the decision, and the hours go before the post does. From there on I’m reasoning for myself. That pattern is why I’d be careful with studies that compare headcounts just before and after a wage rise and find nothing. The margin that moves first is hours and the staffing of the next site, and it moves with a delay.

Why this matters for the thesis

The wage floor rise is usually a response to pain the money system itself created. Living costs rise because the government and central bank aim for prices to rise by about 2% a year, and they create the new money and credit it takes to get there.⁸ In the UK since 2015 the wage floor rise comes from a target set as a share of what the middle earner is paid, not straight from the cost of living.⁹ The Low Pay Commission recommends the floor each year. The money system reaches the floor with a lag, because the middle earner’s pay itself rises as people chase prices. And from the April 2025 recommendations the cost of living sits in the Commission’s instructions by name.¹⁰ Voters demand relief. The relief arrives as a mandated pay rise, which raises the ledger line that brings the machine’s start date forward for the trade. For a single owner short of cash the rise can push their own date back while it raises their reason to buy, and the date then moves at the next site or with the firm that replaces them. And the raise itself is partly an illusion, because if prices rise faster than the pay number, the real wage is falling even as the payslip grows. Inflation is wage deflation viewed from the other side. So the person the policy meant to help gets the automation sooner and keeps less of the raise.

Meanwhile the saving from the screens doesn’t reach anyone as a coffee that costs less than it did. It comes off a price the café’s other costs are pushing up at the same time, so the cup holds or gets dearer, and the expanding money absorbs the gain. Part of it reaches people as higher pay, late and unevenly.¹¹ The rest ends up with whoever is closest to the new money and the assets it lifts. In a system with neutral money [money nobody can expand], the screens arrive on the date the technology alone would have set. That date is later than the policy-loaded one. Policy raises the cost of a person against a machine whose price keeps falling, and that brings the crossover forward. I can’t put a number of years on the gap in general. Under neutral money the screens’ saving lands where it belongs, in the price of the coffee, so the people whose counter hours go face a falling cost of living instead of a rising one while they find the next thing.

The harm is the pairing of the work going sooner than it had to with a rising cost of living. That pairing comes from the design of the money, which is man-made and could be otherwise. Inside the system the interest rate has moved a long way before, in America in the early 1980s, and the central banker who moved it kept the post.¹² The system’s debt then sat mostly with private borrowers. Today the state itself sits on the debt, so the same move would break the system. In Japan prices fell for years at a time with the central bank’s governor staying in post, and the governors through those years were working to end the fall.¹³ Under this design of money, nobody in post has set out to let prices in general fall. In Switzerland, which has twice this century put the design of its money to a national vote, the voters kept what they had both times.

Four arguments against the timing claim

“Automation happens regardless.” True, and I concede it fully. Technology sets the direction, and the destination was already fixed. What policy changes is the timing. Where it raises the cost of employing a person and rivals are free to undercut, it pulls the date in and automation runs beyond its natural pace.

“Higher wages give workers more to spend, which protects jobs.” In the short run the higher pay does get spent, and firms do respond to the extra spending by keeping people on while it lasts. But the standing incentive to automate is what sets the lasting outcome. Once firms have acted on that incentive, the jobs that are left are fewer and different.¹⁴

“Policy doesn’t speed automation everywhere.” Where the state shields a firm from competition or subsidises the old process, it does the opposite and keeps manual roles alive past their natural end. Policy bends the pace away from natural in both directions. It accelerates automation where it raises the wage bill in open competition, and it slows automation where it blocks competition.

“A wage rise only bites like this when the substitute is actually on the shelf.” When the wage floor was new, a rise in it mostly meant higher prices, because there was no kiosk to buy. Today the kiosk and the self-checkout sit in a catalogue with next-day delivery, and the AI that does more and more of the office work is a download away, which is why the wage-to-automation link is tighter now than it was ten years ago.