They’re selling you AI as a helper. A tidy desk assistant that writes emails, summarises meetings, smooths the rough edges off your workload and your mood. It lands as comfort. It also lands as cover. The big move happens underneath the convenience layer, where judgement stops being something companies rent from people and becomes something they own outright. The moment a firm can bottle decision making inside a model, the value of that decision making migrates toward whoever owns the model, the data that feeds it, the compute that runs it, and the distribution that forces everyone else to use it.
That is why the “who cares, it’s just productivity” take is dead on arrival. AI is not hunting jobs by job title. It is hunting repeatable cognition. It comes for the person who spent six years learning to price credit risk, because credit risk is patterns with consequences. It comes for the junior associate who thinks their JD is a forcefield, because a lot of legal work is pattern recognition wrapped in ritual. It comes for the analyst who lives inside spreadsheets, the marketing lead who writes fifteen versions of the same pitch, the project manager who translates humans into templates. The machine does not care about status. It cares about structure. The labour market exposure work makes that blunt. MIT
People keep focusing on whether AI “takes jobs”. That framing misses the mechanism. The mechanism is ownership. Wages have leverage when judgement is scarce and delivered by humans. Wages lose leverage when judgement gets industrialised and bundled into tools you rent access to. You still work, you still “contribute”, you still get told you are “upskilling”. The centre of gravity moves away from labour and toward the owners of the stack.
Now bring in the pipes. The internet already runs on choke points. AI thickens them. Platforms become digital toll booths. Private gravity wells. Everything flows through them because that is where the audience is, where the workflows live, where the distribution sits. They feel personalised and helpful. A casino feels welcoming too. It’s designed to keep you inside, blur time, and make the exit hard to find. You do not leave with more money. You leave trained. That dynamic has had a name for a while. surveillance capitalism The softer version gets sold as modern business. personalisation economy
Here’s the part that should keep you awake. The speed problem.
Technology moves at the speed of light. Policy moves at the speed of paper. People live in the blur between the two.
AI adoption scales in quarters. Labour markets adapt in years. Safety nets adapt in elections. Institutions adapt when forced. The gap produces layoffs and a slow bleed of bargaining power. Then it produces backlash. Not because people suddenly hate technology, because people notice the bargain changed while they were busy paying bills. The modelling keeps pointing at the same pressure line: adoption choices can push returns toward capital and widen inequality. IMF
From here, the system bends into a few recognisable shapes, and the names matter less than the texture. One path keeps markets alive while concentrating power inside the firms that own models, data, compute, and distribution. Daily life becomes optimisation theatre. Another path looks like rent extraction as a lifestyle, where access is the product and dependency is the moat. That techno feudal frame lands because it describes how platforms behave when they become unavoidable. Noema Critics push back and call it monopoly capitalism with sharper language. The incentives still rhyme either way. Jacobin A third path dreams of abundance, where automation makes scarcity feel optional, and politics becomes the only real bottleneck. The hard part is ownership and governance. AI can allocate resources. Someone defines the allocator and owns the rules. Verso
This is where UBI shows up like a friendly plaster. It can be stability. It can also be a collar. A population with weaker bargaining power becomes easier to manage with a monthly payment and a thicker surveillance layer. Calm people subscribe. Calm people do not organise. Calm people keep feeding the engine. The “commons” language keeps appearing because data and infrastructure behave like shared inputs, and ownership treats them like private mines. commons The power problem inside “support” systems is real too. power
AI does not arrive as a robot uprising. It arrives as an accounting improvement. A KPI win. A quiet restructure. A spreadsheet that makes the numbers look better while the ownership layer hardens.
The machine is being built. The deeds are being signed. The toll booths are being installed.
The only question left is simple. Are you a stakeholder, or are you just the data feeding the engine.
If you enjoyed this, drop your email below. No spam, no motivational quotes, no “journey” posts. Just the occasional transmission when the ownership layer shifts again.