Discount intelligence

Unemployed people demonstrating in Helsinki on 24 October 1996. Photo: Seppo Sirkka/Eastpress and Esko Jämsä

So far, the most talked-about AI report of the year has been the “memo from 2028” by the investment research firm Citrini, published a couple of weeks ago. It paints a dystopian picture of the economy in the near future. The report is said to have caused a 224 billion dollar fall on the stock market at the beginning of March.

The scenario goes like this: companies lay off knowledge workers and invest the savings in AI. Intelligence is no longer in short supply. Competition leads to a downward spiral in which companies move more and more of their stakes into AI. Citrini calls this chain reaction the human intelligence displacement spiral. Well-paid white-collar workers are the backbone of consumer spending, and the blow spreads quickly through the whole economy as they go on the dole.

The report goes through how AI could wipe out various business models, hitting travel, insurance, tax, law, SaaS, the platform economy, payments and mortgages, among others:

Booking.com and other travel booking companies. These are the easiest for AI agents to displace. By Q4 2026 an agent would be able to put together a complete trip faster and more cheaply than any platform.

DoorDash, Wolt and the rest of the platform economy. If an AI agent automatically puts every order out to tender, the familiar apps lose their power. In 2027 an agent would choose, out of all the drivers, the one who brings the meal you want to your door for the lowest price. A driver can vibe-code their own app that sells their delivery services to AIs.

Asana, monday.com, ServiceNow and the rest of SaaS. When agentic coding makes building things in-house far easier than before, it pushes SaaS prices down. In 2027 we could use AI to build our own tailor-made Asanas and Mondays.

Mastercard and other payment companies. At the point of payment, a calculating agent would wonder why we are still paying billing fees or intermediary fees? In 2027 agents would start optimising transactions directly and work out how to get around card payments with almost free stablecoins.

Mortgages and bank solvency. The mortgage market rests on the assumption that the borrower stays in work, on at least their current salary, for 30 years. What if they don’t? If the jobs disappear, the mortgage system’s assumption of steady income growth starts to wobble and the bottom falls out of the housing market.

Agents don’t care about brand loyalty or reward points. Wherever there is friction, middlemen or transaction costs, AI pushes the cost down to nearly the price of electricity.

The report’s outcome for 2028 is grim. The S&P 500 has fallen 38% from its peak. Unemployment is over 10%. Labour’s share of GDP has dropped from 56 per cent to 46 in ten years. The whole tax system is built on taxing wages, and when there is less of that work, tax revenue collapses. At the same time the state ought to be expanding safety nets and income transfers, but less money is coming in. Politics would not keep up. AI ought to be taxed, but how, and who gets taxed, would divide opinion. The right would say that taxation reduces growth and competitiveness, and the left would suspect that AI lobbyists will water the whole tax down. An Occupy Silicon Valley movement grows and AI companies would be seen as enemies of the people.

For Finland this scenario would be even harsher. The Nordic welfare model is built on a high employment rate and progressive taxation of earned income. If knowledge work is replaced, the tax base narrows faster than anywhere, because a bigger share of state revenue comes from taxing wages than in the United States, and the public sector’s share is larger. At the same time, the Finnish economy is full of exactly the companies the report refers to: Wolt, Smartly, RELEX Solutions, Supermetrics and dozens of other SaaS and platform companies have been among our few growth stories. If their business models are challenged, Finland loses some of its few international success stories along with them.

Productivity, on the other hand, looks to be growing. Ghost GDP is the report’s new concept, and it is essentially the Solow paradox in new words: productivity shows up in the government’s figures, but the money doesn’t circulate in the real economy in the same way. AI doesn’t need to go for a beer or to the hairdresser after a hard day at work.

Critics (such as Forbes with its one-upping memo from 2030) say, quite fairly, that this won’t happen this fast or this steeply. That may well be. But the markets reacted, which tells you the scenario didn’t feel like it had been plucked out of thin air.

My own addition: if intelligence is no longer a scarce resource, we need to think about what is. For me, it is empathy. The ability to understand another person, their motives, contradictions, fears, desires and cultural context. AI can analyse people better and better, and it is said that to a superintelligent AI we will be like ants. Empathy is still something other than analysis. We can analyse how an ant works, but only another ant can understand another ant. In the end, the value of all services is built on exactly that.