Chapter 3: What Gets Displaced
This is Chapter 3 of 7 in the AI Displacement Series.
Not all work is equally exposed.
This is the most important thing to understand — and the thing most people get wrong.
"AI will change everything" is useless. It doesn't tell you what to do. It doesn't tell you where you're safe and where you're not.
What actually helps is understanding the anatomy of your work. Which parts are exposed. Which parts are protected. And why.
The Anatomy of Knowledge Work
Every knowledge job is made up of tasks. And tasks fall into two categories:
Information work: Gathering, synthesizing, formatting, summarizing, translating, researching, calculating, processing.
Judgment work: Deciding what matters, navigating ambiguity, building relationships, taking responsibility, making calls with incomplete data.
AI is exceptional at information work. It's getting better every month. In many cases, it's already better than humans.
AI is poor at judgment work. It can simulate it. It can assist with it. But it can't own it.
The question isn't "will AI affect my job?" The question is: "What percentage of my job is information work vs. judgment work?"
If your job is 80% information work — gathering data, writing reports, summarizing meetings, formatting documents — you're heavily exposed.
If your job is 80% judgment work — making decisions, building relationships, taking accountability — you're much safer.
Most jobs are somewhere in the middle. The task is figuring out where.
The Exposure Spectrum
The original draft used the illustrative groupings below. It did not supply a dataset, measurement instrument, date or comparable denominator for these percentages. The “high” group even includes 35% and 37% beneath its original 40% threshold. These figures are unverified illustrations, not measured occupation risks, forecasts or an individual job-loss calculator.
Original higher-exposure grouping (unverified percentages):
- Administrative and support roles: 46%
- Legal research and document review: 44%
- Architecture and engineering (routine design): 37%
- Financial analysis and reporting: 35%
- Customer service (scripted interactions): 40%+
Original middle grouping (unverified percentages):
- Marketing (content creation, analysis): 30%
- Software development (routine coding): 25-35%
- Accounting (transaction processing): 30%
- HR (screening, scheduling): 25%
Original lower grouping (unverified percentages):
- Construction and skilled trades: 6%
- Healthcare (direct patient care): 10%
- Sales (relationship-based): 15%
- Executive leadership: 12%
- Skilled negotiation: 10%
But these numbers hide something important: within every job, exposure varies by task.
A lawyer doing document review is heavily exposed. The same lawyer building client relationships and making strategic decisions is not.
A software engineer writing boilerplate code is heavily exposed. The same engineer making architectural decisions and mentoring juniors is not.
A marketer writing blog posts is heavily exposed. The same marketer building brand strategy and cultivating partnerships is not.
The job title doesn't determine exposure. The task mix does.
What I've Watched Disappear
Let me tell you what's happened in my own business over the past two years.
The front-end role.
We had an engineer who'd been with us for two years. Good at their job. Reliable. Part of the team.
In January 2026, we eliminated the role.
Not because of performance. Because of reality.
It happened over the Christmas holidays. I was building things myself — fast, without having to ask anyone. I was picking up front-end knowledge I'd never had before. The work just... got done.
I kept building. And building. And it became clear: the role was no longer needed.
I can do it with Claude Opus now. It wasn't possible before. It wasn't this good before. It was just... good. Now it's actually capable.
That was a hard conversation. Two years together. But the role stopped making sense.
The design agency.
I haven't given our design agency new work in months.
Part of it is slower demand on the consulting side. But part of it is simpler: I didn't use them for my latest website. I built Founder Reality myself.
Their work probably dropped 70-80% from me. Not their fault. Just reality.
The Webflow contractor.
I no longer use Webflow. So our Webflow contractor hasn't gotten gigs from me in a while.
Again — nothing they did wrong. The tool I needed them for became unnecessary.
It's spreading fast.
This isn't one role or one contractor. It's a pattern.
We went from 14 people to 5 to 4. We're staying at 4 for the foreseeable future.
And here's the thing: we're outdoing what used to take 30 people. I can say that confidently.
The old model was: more revenue = more people.
The new model is: more revenue = same people = more freedom.
We're not going back.
The Entry-Level Crisis
The data is stark:
The earlier Stanford Digital Economy Lab analysis reported a 16% relative employment decline for early-career workers (ages 22-25) in the most AI-exposed occupations. This is the historical estimate cited in the original chapter, using ADP payroll records and controls for firm-level changes.
Research context — October 11, 2026. The authors' February 9, 2026 explanation cautions against treating AI as the sole cause. Their August 12, 2026 revision instead emphasizes a 19% employment gap relative to less-exposed peers, using data through June 2026. This is a different measure, described as noncausal, with no evidence of widespread economy-wide job displacement. Neither percentage counts layoffs proven to have been caused by AI or predicts your personal job loss.
Entry-level recruitment is down 13% in jobs exposed to AI since the rise of large language models.
The Big Four accounting firms cut graduate job listings by 44% year-over-year in the UK.
PwC plans to cut graduate hiring by one-third over three years, explicitly citing "the impact of AI."
Law firms are reducing first-year associate classes while keeping overall associate ranks the same.
What's happening is simple: the entry-level tasks are the first to go.
Junior work is heavily information-based. Research. Document prep. Data gathering. Summary writing. Scheduling. Formatting.
All of it is now AI territory.
The ladder is being pulled up. The traditional path — start at the bottom, learn the basics, work your way up — is disappearing.
This doesn't mean there are no jobs. It means the jobs that remain require you to skip the entry-level phase and show up with judgment skills from day one.
That's a fundamental change in how careers work.
The Freelance Collapse
Metric correction — October 11, 2026. The original edition claimed Fiverr downloads fell 18% in the first half of 2024, Upwork downloads fell 22%, and Upwork monthly active users fell 32% over five quarters. The primary dataset, scope and complete periods have not been verified. Those claims cannot establish a collapse in paid freelance work or attribute it to AI.
The companies publish different measures: Fiverr's 2024 results reported annual marketplace buyers down 10% year over year at December 31, 2024, while Upwork's 2024 results reported active clients down 2% at that date. Buyer and client counts are not app downloads, monthly active users, freelancer earnings or employment, and these figures do not establish AI causality.
The lists below express my interpretation of task pressure at the time, rather than outcomes established by those company metrics.
What's being displaced:
- Basic writing and copywriting
- Simple graphic design
- Data entry and formatting
- Translation
- Basic coding tasks
- Research and summarization
What's surviving:
- Complex creative direction
- Strategy and consulting
- Relationship-based work
- Tasks requiring deep domain expertise
- Work that requires accountability
The pattern is the same: information work is collapsing. Judgment work is holding.
Clients expect lower prices because AI makes tasks cheaper and faster. Freelancers face pressure to lower rates or lose work to AI alternatives.
Some freelancers are pivoting — positioning themselves as people who manage and direct AI tools rather than doing the raw execution. That's the adaptation.
A Framework for Your Own Role
Here's how to analyze your exposure:
Step 1: List your tasks.
Not your job description. Your actual tasks. What do you spend time on in a typical week?
Be specific. Not "marketing" — "writing blog posts," "analyzing campaign performance," "building influencer relationships," "presenting to stakeholders."
Step 2: Categorize each task.
For each task, ask: Is this primarily information work or judgment work?
- Information work: Could AI do this if given the right inputs? Is it about gathering, processing, formatting, or translating information?
- Judgment work: Does this require decisions that can't be easily codified? Does it require relationships? Does it require accountability?
Be honest. Most people overestimate how much judgment their work requires.
Step 3: Calculate your exposure.
Add up the time you spend on information work vs. judgment work.
If you're 70% information work, you're heavily exposed. If you're 50/50, you're moderately exposed. If you're 70% judgment work, you're relatively protected.
Step 4: Look at the trend.
Which of your information tasks are already being done by AI somewhere? Which will be in 12 months?
The exposure isn't static. It's increasing.
The Tasks AI Does Well
Let me be specific about what AI handles now:
Research and synthesis
- Finding information across sources
- Summarizing documents and meetings
- Comparing options and creating analyses
- Answering factual questions
Content creation
- Writing first drafts
- Editing and formatting
- Translation
- Code generation
- Design iteration
Data processing
- Categorization and tagging
- Pattern recognition
- Report generation
- Calculation and analysis
Coordination
- Scheduling and logistics
- Status tracking
- Routing and triage
- Template-based communication
If your job is primarily these tasks, you're competing with AI directly.
The Tasks AI Does Poorly
And here's what AI still struggles with:
Novel judgment
- Decisions without clear precedent
- Trade-offs that require values, not just data
- Calls that could go either way
Relationship navigation
- Building trust over time
- Reading between the lines
- Managing sensitive interpersonal dynamics
- Knowing when to push and when to back off
Accountability
- Taking responsibility for outcomes
- Being the person who gets blamed if it goes wrong
- Making the call when no one else will
Physical presence
- Being in the room
- Hands-on execution
- Sensory judgment
Taste and selection
- Knowing what matters
- Choosing what to pursue and what to ignore
- Curating and editing with a point of view
If your job is primarily these tasks, you're relatively protected.
The Shift Within Roles
Here's what's actually happening: roles aren't disappearing. They're compressing.
The junior associate used to do 80% information work and 20% judgment work. Now the 80% is handled by AI. The role either disappears or becomes 100% judgment work from day one.
The marketing coordinator used to spend half their time on admin and logistics. Now AI handles that. They either move up to strategy or they're redundant.
The software engineer used to write boilerplate code for hours. Now AI generates it in seconds. They either move to architecture and review or they're competing with a tool.
The pattern everywhere: the information layer is collapsing into AI. The judgment layer is what remains.
This means fewer people doing more judgment work. It means entry-level disappears. It means mid-level gets squeezed.
It also means opportunity — for people who can operate in the judgment layer.
What This Means For You
If you're heavily exposed:
Start moving now. The longer you wait, the more compressed the timeline becomes. Don't wait for the layoff.
Build skills in the protected tasks. Relationship building. Decision making. Accountability. These don't come naturally to everyone, but they can be developed.
Position yourself as the person who directs AI, not the person AI replaces. Learn to use the tools. Become the human in the loop.
If you're moderately exposed:
Understand which tasks are at risk and which aren't. Be intentional about spending more time in protected territory.
Prepare for your role to compress. What remains after AI takes the information layer? Make sure you can do that.
Build options outside your current job. The role might change faster than you expect.
If you're relatively protected:
Don't get complacent. "Protected" doesn't mean "immune." The boundary moves every month.
Build skills that compound. Judgment gets better with practice. Relationships deepen over time. These are assets.
Help others adapt. You're in a position of relative strength. Use it.
What's Next
This is Chapter 3 of 7.
Now you understand what's actually getting displaced — information work, entry-level tasks, the execution layer.
But understanding what's exposed isn't enough. You need to know where to go.
In Chapter 4, I'm going to show you the five layers AI can't eat. Not vague "soft skills" — a specific framework for positioning yourself in the parts of work that will always require humans.
This is where it gets practical.