Saurav Das  /  Writing
Economic Analysis  ·  2026
Special Report

The Economics of a
Post-AI World

AI is reshaping the workforce faster than policy can respond. The data is here. The question is whether governments, institutions, and societies will act before the default outcome gets locked in.

By Saurav Das  ·  2026

We are getting that dopamine hit every time an AI agent does something for us autonomously—drafting a report, writing code, managing a workflow. It feels very sci-fi. But sci-fi has a habit of dragging dystopia in right behind it. The movie Elysium gave us a useful mental model: a world split by access, where a large part of society is locked out of basic needs while a small minority lives in abundance. That felt fictional in 2013. It reads more like a forecast today.

The question is no longer whether AI will displace significant numbers of workers. The data is in. The question is whether public institutions will design a response before the damage is done—and whether the upside will be broadly shared or concentrated in the hands of those who own the infrastructure.

"The hard part of the equation isn't just the models. It's the economics and society. It's governance. It's distribution." — Demis Hassabis, CEO, Google DeepMind
Figure 1
The Scale of Disruption at a Glance
Sources: WEF 2025 · Goldman Sachs 2025 · McKinsey 2025
92M
Jobs projected to be displaced globally by 2030
WEF Future of Jobs 2025
170M
New roles expected to emerge by 2030
WEF Future of Jobs 2025
57%
U.S. work hours automatable with today's technology
McKinsey 2025
47%
Industry tasks where AI matches or outperforms professionals
Goldman Sachs 2025
50%
Entry-level white-collar jobs at risk within 5 years
Dario Amodei, Anthropic
15%
Projected U.S. productivity gain when AI is fully integrated
Goldman Sachs Research
Macro projections show net job growth — but the distribution of displacement is highly uneven across age, sector, and income level.
01 —

The Data on Disruption

The aggregate employment picture looks relatively stable. The ground-level picture is not. Macro forecasts from the World Economic Forum project 92 million roles displaced by 2030 alongside 170 million new ones—a net gain of 78 million on paper. But those averages obscure the distribution. McKinsey estimates that today's existing technology could already automate approximately 57% of current U.S. work hours if fully deployed.

The pressure is arriving first on the youngest workers. Goldman Sachs data from 2025 shows unemployment among 20–30-year-olds in tech-exposed occupations has risen nearly 3 percentage points since early 2025—the steepest increase in youth tech unemployment in over a decade.

Figure 2
Global Job Displacement vs. Creation by 2030
WEF Future of Jobs Report 2025

Each square = ~1 million jobs. 262 total squares shown (proportional).

92M Displaced
92M Replaced (equivalent)
+78M Net New Jobs
Stable
Net gain of 78 million looks encouraging — until you account for skill mismatch, geography, and the 5–10 year transition lag that workers in displaced roles must navigate.
Figure 3
AI Automation Exposure by Sector (%)
Goldman Sachs · SQ Magazine · McKinsey 2025
Administrative / Office
78%
Customer Service
65%
Legal / Paralegal
44%
Software / Programming
40%
Data Processing / Analytics
65%
Production / Manufacturing
26%
Healthcare / Caregiving
12%
Construction / Trades
6%
Knowledge work and repetitive cognitive tasks face the highest exposure. Physical, dexterous, and relationship-intensive work remains comparatively insulated.
Figure 4
The Early-Career Gap: Employment Decline in AI-Exposed Roles (2022–2025)
Goldman Sachs Research 2025 · Federal Reserve Bank of NY
Software developers, age 22–25
−20%
All workers age 20–30, tech-exposed roles
−3 pts
Recent college graduates (NY Fed)
5.8% UR
Big Tech new grad hiring decline (2024 vs 2023)
−25%
Overall employment (all workers)
+Stable
Key insight: AI appears to be suppressing hiring more than destroying existing jobs in the near term — a pattern Goldman Sachs describes as employers using AI to avoid adding headcount rather than immediately firing existing workers. The downstream consequences for a generation unable to find entry-level positions may be as severe as direct displacement — just slower and harder to measure.
Overall employment remains stable, masking a pronounced structural shift at the entry level where AI is most directly competing with new workers.

02 —

Two Hard Scenarios

The macro forecasts converge on two difficult outcomes that aren't mutually exclusive. The disruption of knowledge work is already happening. The inequality question is structural.

Figure 5
The K-Shaped Economy: Two Trajectories
Goldman Sachs · McKinsey · Geoffrey Hinton (BBC)

⬆ Capital Owner Trajectory

  • AI infrastructure owners capture productivity surplus
  • Returns on capital outpace returns on labor
  • A few tech giants capture disproportionate gains
  • Wealth concentration accelerates
  • Hyperscaler capex hit $142B/quarter in late 2025
  • S&P 500 buybacks topped $1.02T in past 12 months

⬇ Wage Earner Trajectory

  • Entry-level positions compressed or eliminated
  • Bargaining power eroded by AI substitution
  • Upskilling demand exceeds retraining supply
  • Geographic concentration of AI gains in few metros
  • 40% of white-collar job seekers failed to get interviews (2024)
  • Hiring for $96K+ roles at decade-low
Geoffrey Hinton: "Most of the financial gains will go to the rich and not the people whose jobs are lost, and that's going to be very bad for society."
Figure 6
Documented AI-Attributed Job Losses: Exponential Trajectory (2023–2025)
Challenger, Gray & Christmas · Medium / Steve Kaplan AI Research
0 25K 50K 75K 100K 45K 65K 78K ~110K? 2023 2024 2025 2026 proj. Documented AI layoffs Projection (extrapolated)
AI-attributed layoffs have followed an exponential trajectory. McKinsey notes that only 1% of companies believe they are at AI implementation maturity — meaning current figures likely represent the early slope, not the peak.

03 —

The Policy Response

The good news is that the conversation has moved from theoretical to legislative. Several concrete mechanisms are now under active development at the federal, state, and international level.

The fiscal math is challenging but tractable for targeted versions. A poverty-line UBI for all U.S. adults would cost approximately $8.5 trillion annually against $4.9 trillion in federal revenue — clearly infeasible in full. But targeted guaranteed income pilots, automation taxes, and sovereign wealth fund models are already being tested.

Figure 7
Policy Mechanisms on the Table: Status Overview
GovFacts 2025 · LSE Business Review · Newsweek · Wikipedia
Mechanism Key Proposal Scale Status
Guaranteed Income Pilot H.R. 5830 — $495M/yr federal pilot, 5 years National pilot Proposed
American Equity Fund Sam Altman — AI companies contribute 2.5% value/yr into citizen dividend National Proposed
Robot / Automation Tax Bill Gates (2017) — levy on AI productivity gains comparable to displaced worker taxes National Proposed
Alaska Permanent Fund ~$1,600/resident annually from sovereign wealth fund State (AK) Active
AI Training Grants (U.S. DOL) $30M in AI literacy + $98M pre-apprenticeship grants Federal Active
Marshall Islands UBI ~$200/quarter to every citizen from trust fund National Active (Nov 2025)
Germany AI Investment €1B public funding for AI research and skills National Active
Singapore AI Tax Deduction 400% deduction on AI expenses up to S$50,000 National Active
NYC Guaranteed Income $1,000–$2,500/month for homeless youth and pregnant mothers Municipal Pilot
Data Royalty Model AI companies pay royalty on training data; proceeds to sovereign wealth fund National Theoretical
As of 2026, no country has implemented a full UBI system. The policy response lags the technology by a significant margin.
Figure 8
The Funding Gap: What Would UBI Cost vs. What Exists?
Tax Project Institute · Newsweek · U.S. Treasury
U.S. Federal Revenue (2024)
$4.9T
S&P 500 Buybacks (past 12 months)
$1.02T
Hyperscaler AI capex (Q4 2025, annualized)
$568B
Yang $1K/month UBI (all adults) annual cost
$2.8–3T
UBI at poverty line for all U.S. adults
$8.5T
H.R. 5830 Guaranteed Income Pilot (annual)
$495M
The redistribution lens: Redirecting even a fraction of hyperscaler capex ($568B/yr annualized) or corporate buybacks ($1.02T) into a sovereign wealth fund would fund meaningful targeted income support without requiring new tax revenue — structurally similar to the Alaska Permanent Fund model.
Full universal UBI is fiscally immense. But targeted, pilot-scale programs and fund-based redistribution mechanisms are well within reach of existing capital flows in the AI economy.

04 —

The Institutional Gap

The most alarming part of this moment isn't the technology — it's the governance vacuum. Both Dario Amodei and Demis Hassabis stated plainly that governments are not sufficiently prepared to address the economic and societal changes that advanced AI brings. Hassabis expressed genuine surprise that more professional economists are not seriously modeling what a post-AGI world looks like.

"Do we have the right institutions to distribute this new productivity, this new wealth, more fairly?" — Demis Hassabis, on post-AGI governance
Figure 9
Institutional Response vs. Pace of AI Disruption
McKinsey · WEF · Hassabis/Amodei conversation (2025)
Pace of AI Disruption
🔴 AI coding benchmark: 4.4% → 71.7% task completion in 12 months (SWE-Bench)
🔴 GPT-class models now in top 10% on bar exam simulations
🔴 AI adoption: 71% of orgs using AI in at least one function (McKinsey 2025)
🔴 Only 5.4% of firms have formally adopted GenAI — meaning acceleration is early-stage
Pace of Policy Response
🟡 No country has implemented full UBI as of 2026
🟡 DOL AI literacy program: $30M — vs. $568B/yr in AI capex
🟡 H.R. 5830 still in proposal stage; $495M pilot not yet funded
🟡 Google DeepMind just hired its first Director of AGI Economics — a role most governments haven't created yet
The asymmetry is stark: AI capability is scaling exponentially while governance and redistribution infrastructure is advancing incrementally. This gap is the core risk.

05 —

The Forward View: What Needs to Happen

There is a version of this future that is genuinely good. AI-driven productivity could fund better healthcare, stronger infrastructure, new scientific breakthroughs, and broader access to opportunity. The WEF projects 170 million new jobs alongside 92 million displaced. Goldman Sachs models suggest transitional unemployment effects are historically short-lived, resolving within two years as new roles emerge.

But "net positive" and "equitably distributed" are not the same thing. The design choices made in the next 3–5 years will determine whether AI's productivity gains are broadly shared or captured entirely by the owners of the infrastructure. The default, without intentional design, is concentration.

Figure 10
Priority Actions: Closing the Governance Gap
Synthesis: Hassabis · Amodei · WEF · LSE
01
Establish a national AI productivity dividend mechanism
Model: Alaska Permanent Fund applied to AI infrastructure revenues. Even a 1% annual contribution from hyperscaler capex ($568B) would generate $5.7B/year for redistribution.
02
Fund guaranteed income pilots at scale, not study scale
H.R. 5830's $495M is a start. Finland's 2017 trial improved wellbeing. A 2024 U.S. study showed no meaningful labor supply reduction at $1K/month. Expand the evidence base before the displacement wave arrives.
03
Treat training data as a public resource
AI is trained on the collective knowledge of humanity. Royalty structures on training data — analogous to oil and mineral extraction fees — would create a legitimate claim for public dividends without requiring new tax structures.
04
Build post-AGI economics as a discipline, not an afterthought
Google DeepMind just hired its first Director of AGI Economics. Most governments, universities, and multilateral institutions haven't. McKinsey notes 14% of the global workforce — 375 million people — will need career transitions by 2030. That requires institutional infrastructure built now.
The goal is not to slow AI progress. The goal is to build the economic and policy architecture in parallel with the technology — before the default outcome is locked in.
"AI could double or triple productivity, but if we don't redesign how gains are distributed, we'll see even more concentration of wealth and power." — Scott Santens, UBI Advocate