Where the jobs go. And why Elon keeps saying UBI.

Two signals from the Bank of Canada in the same month. Aave got a formal paper calling it a functioning non-bank lender with zero bad debt. Open banking got a press statement saying a 2026 launch would be premature. Most people read those as separate stories. They are the same story at different speeds.

Zero margin finance works. Public data rails are coming. Apply both to anything a human does on a screen. Then notice where Elon Musk, Sam Altman, and half the frontier tech class keep parking the conversation. Universal basic income. Over and over. That is not noise. That is people who can see the next curve telling you what they already priced in.

This is not a hypothetical anymore. The numbers are already on the tape.

The layoff tape is already running.

Employers directly linked 55,000 job cuts to AI in 2025. That is 12 times the number from two years earlier. Modeling that includes unattributed cases puts the actual AI displaced or AI foregone positions between 200,000 and 300,000 in 2025 alone.

In the first three months of 2026 the tech sector cut 45,000 jobs. Roughly 20 percent were explicitly tied to AI and automation. Anthropic CEO Dario Amodei said in 2025 that AI could eliminate around 50 percent of entry level white collar positions inside five years. CEOs at Ford, Amazon, Salesforce, and JP Morgan have publicly said large white collar cuts are coming.

The World Economic Forum projects 92 million jobs displaced by 2030 and 170 million new ones. Net positive on paper. But the new jobs concentrate in major metros and require credentials most displaced administrative workers do not have today. That gap is the story.

If a screen runs the work, an agent runs the work.

Look at what a financial advisor does. Open an account, portfolio checks, product explanation, KYC paperwork, underwriting against known rules. All of it runs on a screen.

Now add an agent with open banking read and write access. It sees every account across every institution in real time, pulls the right data, matches the rules, and pushes a decision through a bank app at 2 AM on a Sunday. Not a chatbot. Actual work. Without a human in the loop.

Anthropic published its Economic Index in January 2026. In their sample, 36 percent of jobs had Claude being used for at least a quarter of their tasks. Pooled across reports that number climbed to 49 percent. Business and finance occupations show 94.3 percent theoretical AI coverage and 28.4 percent observed usage today. Those are not projections. Those are logs of real work already getting done by a model.

The question is not whether an agent can do the work. The question is how many human seats are still required once the agent can pull the right data and act on it.

Where five desks become one or two.

Wolters Kluwer found 44 percent of finance teams will be running agentic AI in 2026. That is a 600 percent jump. In wealth management specifically, agentic AI is already cutting advisor time on manual prospecting by 40 to 50 percent, increasing net new AUM by 30 to 40 percent, and reducing onboarding costs by 30 to 40 percent while accelerating onboarding by 50 percent.

Compound that for 24 months and the headcount math writes itself. The compression is not five to zero. It is five to one or two. The routine 80 percent collapses. The judgment layer survives. Complex estate planning, cross border tax, succession, stress cases. Too much context for an agent to close.

Same pattern in underwriting. Same pattern in onboarding. Same pattern in compliance review. Same pattern in research. Any knowledge work role where the deliverable lives on a screen follows the same curve.

Orchestration is the forcing function.

The next phase of the AI cycle is not one bigger language model. It is orchestration. Agents calling agents. Agents handing off to agents. Agents with tools and memory running loops.

Agents are the operating system for AI models. They do not just talk. They do. They pull data. They write data. They trigger workflows. They hand a result to the next agent in the chain and wait for a response. According to a June 2025 estimate, the ability of agents to automate tasks was already doubling every three to seven months.

At the hardware layer, NVIDIA shipped Isaac GR00T N1 as the first open foundation model for humanoid robots in 2025 and pushed out N1.7 as a vision language action model for generalized robot skills. At the protocol layer, Aave crossed 1 trillion dollars in cumulative lending volume on February 25, 2026. That is one smart contract running at roughly zero margin doing more lending in aggregate than most countries have GDP for.

This is not a 2028 story. This is shipped and running in production right now.

Why Elon keeps saying UBI.

Musk has been talking universal basic income since 2016. Andrew Yang ran a presidential campaign on it. Sam Altman funded a three year randomized controlled UBI study through OpenResearch, giving 1,000 people in Texas and Illinois 1,000 dollars a month and comparing them to a 2,000 person control group getting 50 dollars a month.

The results published in 2024 are the part most people miss. Recipients worked about 1.3 fewer hours per week and were 2 percentage points less likely to be employed. That is a labor supply shift you can measure. They also reported reduced stress and food insecurity in year one, and increased agency. More likely to start a business. More likely to take a lower paying job for more independence. More likely to budget and plan ahead.

People inside the frontier are not treating job compression as a speculative topic. They are treating it as a known shape and running trials on the response. The UBI talk is the forecast leaking out.

You do not have to agree with UBI as the answer. You have to notice that the people closest to the machine already assume the compression is coming.

The blocker is cultural, not technical.

The tech is shipped. That is not the blocker. The blocker is an institutional posture that treats tech adoption as optional, or worse, as taboo. Forms still filled like a human reads them line by line. Reviews still run as if volume is manageable with more headcount. Rules engines from 2017. An estimated 9 million Canadians still use screen scraping to link bank accounts, despite the security risk, because the public rail is late.

Time is moving faster than a laid back corporate posture can absorb. Treating technology adoption as taboo is a decision to be the one seat out of five that gets kept, not one of the other four.

What to do with an 18 to 24 month window.

If your job is the kind that could compress, three moves.

One. Learn what an open banking API actually does. The endpoints, the scopes, the read and write boundaries. The UK version already has 12 million active users and has added roughly 4 billion pounds to the UK economy. The Canadian rail is arriving whether the Bank of Canada hits a date or not.

Two. Learn what an agent actually is. Not as a buzzword. As a loop with tools, memory, and the ability to call other agents. Build one yourself if you can. Even a small one.

Three. Pick a judgment heavy specialty that does not compress. Estate, cross border, stress cases, regulatory edge cases. Build deep stack there.

Margin compression is slow at first. Then it is not. The UBI chatter is not a political preference. It is a forecast leaking out of the people closest to the machine. The layoff numbers are already on the tape. The Anthropic Economic Index is the receipt. The wealth management efficiency numbers are the preview.

Read the signals or be one of them.

Sources and data.

Layoffs and labor data: Fortune on CFO survey Mar 2026, HBR on AI layoffs Jan 2026, CBS News on AI layoffs 2026. Anthropic Economic Index: January 2026 report. Finance team agentic AI adoption: Wolters Kluwer via Neurons Lab. Wealth management efficiency: Cognizant 2026 predictions. Altman UBI study: CBS News summary, BIEN analysis. Aave and DeFi: DefiLlama Aave, CoinLaw Aave statistics 2026. NVIDIA GR00T: NVIDIA Newsroom GR00T N1. Open banking Canada: C.D. Howe Institute, Open Banking Tracker. Goldman Sachs: AI labor market analysis. WEF workforce projection: referenced via ALM Corp displacement statistics.