← → Navigate or scroll right

When the rules change

AI, work, and who captures the benefit

This is a discussion, not a presentation.
Ask questions. Challenge ideas. Share perspectives.

A question I keep hearing

At almost every Future Together meetup,
someone asks a version of the same thing:

“How do I make sure
I’m not replaced by AI?”

It’s the right question.
But it’s not the only one worth asking.

What are you seeing in your work or industry right now?

What the data actually shows

Stanford researchers tracked 25 million US workers
using ADP payroll records through September 2025.¹

Headline unemployment: 4.9% — barely changed.

Workers aged 22–25 in AI-exposed jobs — coding, writing, legal, customer service — saw employment drop by ~20% since late 2022.¹

Workers in the same roles, aged 35+: up 8%+
The pattern begins almost exactly when ChatGPT launched.

Independently confirmed: Anthropic’s March 2026 research — using their own usage data, not third-party payroll records — reaches the same conclusion.

¹ Canaries in the Coal Mine: AI’s Early Labour Market Signal (Stanford Digital Economy Lab, Nov 2025) — digitaleconomy.stanford.edu

Who’s actually most exposed?

The answer from Anthropic’s own research isn’t what most people expect.

Most exposed workers

Older. More educated. Higher paid.
More likely to be women.

Not the entry-level picture most people assume.

Most exposed occupations

💻 Computer programmers — 75% of tasks
📞 Customer service reps — heavy automation
💰 Financial analysts — significant coverage
📄 Data entry keyers — most tasks automated

AI is currently at ~33% of its theoretical capability in technical roles.
That gap will close.

Does this change how you’re thinking about your own situation?

Anthropic: Labor Market Impacts of AI: A New Measure and Early Evidence (Mar 2026) — anthropic.com/research/labor-market-impacts

Why the headline numbers mislead

“But unemployment is stable — isn’t that evidence
AI isn’t having an impact?”

📊 Wrong metric

Unemployment counts only those actively seeking work. Not those who’ve given up. Not those reduced from 38 to 15 hours. Underemployment is the more honest signal.

⏳ Wrong timing

Firms absorb new capability before restructuring. A firm that can do 90% of its work with AI hires offshore staff first. The structural reckoning comes later.

🚪 Wrong door

Displacement appears first in who isn’t hired, not who’s fired. Graduate intake freezes. Entry-level roles disappear quietly.

📉 Lagging indicator

Unemployment data follows economic reality by months or years. By the time it moves, the structural shift has already compounded.

What signals should we be watching?

Not all AI use is equal

The Stanford paper draws a crucial distinction:

Automative

AI does the task instead of the human.

Result: employment and wages decline in affected roles.

Augmentative

AI helps the human think, check, or produce better work.

Result: employment and wages hold or grow.

The data shows automation happening first
where work is most codifiable — entry-level, routine, documented.

In your workplace: is AI being used to replace tasks, or to help people do more?

Why experience protects you — for now

Codified knowledge

Documented processes. Established workflows. Things that can be written down and handed to a system.

AI is competitive with this.

Mostly held by entry-level workers.

Tacit knowledge

Judgment built over years. Client relationships. Knowing what to do in the situation that doesn’t fit the pattern.

AI complements this.

Mostly held by experienced workers.

Experience provides a buffer. For now.
But tacit knowledge is developed through the entry-level years.

The missing rung

If AI handles what juniors used to do,
the pipeline for experienced workers is being disrupted at its source.

If there are no entry-level roles,
where does the next generation of senior talent come from?

This is already a live question in law, accounting,
consulting, and software development.

Who in your field trained you? How would someone learn that now?

A two-tier economy

If the trajectory continues without intervention:

Group 1 — smaller

Skills that complement AI systems.

Judgment. Relationships. Accountability. Taste.

Economic participation: valued and growing.

Group 2 — larger

Skills that AI can substitute.

Codifiable, documented, routine.

Economic participation: structurally optional.

That’s not a stable arrangement.
History is reasonably clear on what happens next.

Where do the people you care about sit in this picture?

Skills that may provide protection

Discussed by academics and practitioners working closely with the transition:

🔍 Attention to detail

Catching what AI misses. Knowing when output is wrong even when it looks right.

⚙️ Managing AI agents

Knowing how to direct, verify, and steer AI systems effectively. A new and learnable skill.

🎨 Exceptional taste

The ability to judge quality and fit in ways that are hard to define but easy to recognise.

🤝 A trusted network

Relationships where people trust you specifically — not just any competent source.

Honest caveat: even these aren’t guaranteed.
Predicting individual outcomes is like forecasting tomorrow’s weather.
The climate is more predictable than the weather.

Which of these feels most achievable for you right now?

The Monopoly problem

Imagine you’re playing Monopoly. One player owns most of the board.

Then that player discovers they no longer need
the other players to land on their properties to collect rent.

The income just arrives anyway.

Our economic system is built on a foundational assumption:
human labour is a necessary input to production.

For the first time in history, that assumption may be weakening.

If AI does most of the productive work, who gets the benefit — and how?

Two possible futures

🌟 The optimistic path

AI generates so much productivity that society becomes genuinely wealthier — broadly, not just at the top.

Shorter working weeks. New forms of social income. More time for family, community, care, and meaning.

A real possibility. Worth working toward.

⚠️ The drift path

No catastrophe. Just drift.

A slow narrowing of who the economy is really for. Two tiers. Declining middle. Social unrest arrives when enough people realise what has happened.

Not inevitable. But the current direction.

The optimistic future doesn’t arrive automatically.
It requires deliberate choices — by individuals, organisations, and governments.

What would need to change for us to take the optimistic path?

Not panic. Not paralysis. Action.

The gap between what’s actually happening
and what most people understand is enormous.

💡

Awareness

Understand what’s actually happening. Separate hype from signal.

→

💬

Conversation

Talk to people. Challenge ideas. Bring others into the picture. This is where it starts.

→

🚀

Action

Informed, grounded, community-supported steps forward. Not heroics. Just motion.

You don’t have to figure this out alone.

Who in your life needs to be part of this conversation?

Resources & Discussion

Scan QR codes to access resources

📊 Canaries in the Coal Mine

Stanford Digital Economy Lab — the 25 million worker study

Stanford paper (PDF)

📝 Read the full article

When the rules change: AI, work, and who captures the benefit

futuretogether.community/blog

🌐 Join Future Together

Be part of the ongoing conversation. Monthly meetups. Community.

futuretogether.community/join

🔬 Anthropic Labor Market Research

A new measure of AI exposure — using actual usage data, not forecasts

anthropic.com/research

“The future is arriving. Let’s face it together.”

Let’s continue the conversation.

futuretogether.community

Created with Beyond Better