What Is the Anthropic AI Exposure Index? How to Find Out If Your Job Is at Risk
Anthropic's AI Exposure Index maps 800+ occupations against real Claude usage data. Here's how to read it and what it means for your career.
Why This Index Is Different From Every Other “AI Jobs” Study
Most reports about AI and employment rely on the same approach: take a list of job tasks, ask experts whether AI could theoretically perform them, and rank occupations by exposure. The methodology sounds reasonable until you realize it’s based entirely on speculation about what AI might do rather than what it’s actually doing.
The Anthropic AI Exposure Index — formally published as the Anthropic Economic Index — takes a different approach. Instead of asking what’s theoretically possible, it looks at what’s actually happening: real conversations between real users and Claude, analyzed at scale to understand which job tasks AI is already being used for. That distinction matters more than it might seem.
This article explains exactly what the Anthropic Economic Index measures, how its methodology works, which occupations it flags as most and least exposed, and how you can use this information to think clearly about your own career.
What the Anthropic Economic Index Actually Measures
The Anthropic Economic Index is research published by Anthropic in early 2025. It maps AI usage patterns drawn from Claude interactions against the Department of Labor’s O*NET database, which catalogs the tasks and activities that define over 900 occupations.
The core question it tries to answer: Which parts of which jobs are people already using AI to help with?
This is narrower — and more meaningful — than asking which jobs AI could eventually replace. The index doesn’t model future scenarios. It describes the present. That makes it a useful reality check against the more speculative analyses that often dominate headlines.
How the Data Was Collected
Anthropic sampled a large volume of anonymized Claude conversations — across both API usage by developers and consumer/business usage — and ran each conversation through a classification pipeline. That pipeline identified what task was being performed in each conversation.
Those tasks were then matched to the standardized task descriptions in ONET. ONET assigns each occupation a set of activities and skill requirements, so once a task is matched to O*NET’s taxonomy, it can be linked to the occupations that involve that task.
The result is an exposure score for each occupation: roughly, what proportion of the tasks that define that job are tasks people are already using Claude to assist with.
Two Types of Exposure: Augmentation vs. Automation
One of the most important distinctions the research makes is between augmentation and automation.
- Augmentation: AI helps a human complete a task faster or better. The human is still doing the work — AI is a tool within their workflow.
- Automation: AI completes a task with minimal or no human involvement. The task is delegated to the AI.
The research found that the majority of current Claude usage falls into the augmentation category. People are using AI to help them write, analyze, debug, and research — not to fully hand off those activities. This doesn’t mean automation never happens, but the current pattern skews heavily toward AI as assistant rather than AI as replacement.
That finding matters a lot for how you interpret a high exposure score. High exposure doesn’t automatically mean high displacement risk. It can just as easily mean high opportunity for productivity gains.
Which Occupations Show the Highest Exposure
The Anthropic Economic Index identifies a clear pattern: occupations requiring complex knowledge work, particularly in STEM, writing-intensive, and analytical domains, show the highest exposure scores.
Top Exposed Occupation Categories
Computer and Mathematical Occupations Software developers, data scientists, statisticians, and mathematicians rank among the highest-exposure jobs in the index. This tracks with common sense — these are the workers most likely to be using Claude for code generation, data analysis, documentation, and algorithm development. The exposure here is very real, but current usage suggests augmentation dominates: developers use AI to write boilerplate, debug code, and draft comments, not to hand over entire projects.
Legal and Business Analysis Paralegals, legal researchers, compliance analysts, and business intelligence analysts show high exposure. Document review, contract summarization, research synthesis, and report drafting are tasks Claude handles well, and these are core activities in those jobs.
Writing, Media, and Content Technical writers, copywriters, journalists, and content strategists see high exposure across the board. AI’s competency in drafting, editing, and researching written content directly overlaps with a significant share of these roles’ responsibilities.
Financial Analysis Financial analysts, accountants working on reporting, and actuaries show elevated exposure, particularly for tasks involving data interpretation, scenario modeling, and documentation.
Occupations With Lower Exposure
Jobs requiring physical presence, hands-on skill, or real-time situational judgment tend to score lower.
- Trades and construction: Electricians, plumbers, HVAC technicians — the majority of their work can’t be delegated to a text-based AI.
- Healthcare (hands-on): Nurses, physical therapists, dental hygienists — patient care involves physical skill and contextual judgment that isn’t easily replaced.
- Food service and hospitality: The work is inherently physical and interpersonal.
- Emergency services: Firefighters, paramedics — field decisions in dynamic environments remain far outside AI’s current reach.
The Income Gradient
One of the more counterintuitive findings: higher-earning occupations tend to show more AI exposure, not less. Previous automation waves primarily displaced lower-wage, repetitive physical tasks. AI appears to be different — it’s targeting the knowledge work that historically insulated workers from automation. White-collar, high-credentialed jobs show more overlap with what AI does well than blue-collar, trade, or service roles.
This doesn’t mean high-earners are doomed. It means the impact of AI on knowledge workers is likely to be significant, and it’s happening now, not in some abstract future.
How to Look Up Your Own Occupation
The Anthropic Economic Index published its findings with enough occupational granularity that you can identify roughly where your own job falls.
Step 1: Find Your O*NET Occupation
The Department of Labor’s O*NET OnLine database lets you search by job title and returns a standardized occupational code. Look up your current role and note the tasks and activities listed for it.
Step 2: Cross-Reference With the Index’s Categories
The Anthropic Economic Index organizes results by broad O*NET occupational groups. Find which group your job falls into:
- Management occupations
- Computer and mathematical occupations
- Legal occupations
- Business and financial operations
- Arts, design, entertainment, and media
- Healthcare (clinical vs. administrative)
- Installation, maintenance, and repair
- Production occupations
Each of these groups has a corresponding exposure profile in the research.
Step 3: Think at the Task Level, Not the Job Level
The most useful insight from the index is that exposure is measured at the task level, not the job level. Almost every job has a mix of high-exposure and low-exposure tasks. A marketing manager might spend 30% of their time on AI-augmentable work (drafting briefs, summarizing reports) and 70% on lower-exposure work (client relationships, strategic decisions, team management).
Rather than asking “Is my job at risk?” ask: “Which of my tasks overlap most with what AI is good at, and what does that mean for how I should spend my time?”
What “High Exposure” Should Actually Change About How You Work
A high exposure score isn’t a warning to polish your resume. It’s information about where efficiency gains are available — and where competitive advantage is shifting.
If Your Job Is High-Exposure
Workers in high-exposure occupations who are already using AI tools effectively are pulling ahead. The gap between someone who uses Claude well and someone who doesn’t is widening. High exposure means your core tasks are exactly the ones AI can accelerate — and that’s a competitive advantage if you’re using it.
The risk isn’t AI replacing you. The more immediate risk is a colleague or competitor doing the same work faster because they’re using AI and you aren’t.
If Your Job Is Lower-Exposure
Lower-exposure jobs aren’t immune — they often contain a subset of tasks (documentation, scheduling, communication) that are highly automatable even if the core work isn’t. Focusing AI adoption on those peripheral tasks can free up significant time for the hands-on work that defines the role.
The Hybrid Reality
Most knowledge workers will end up in a hybrid pattern: some tasks offloaded to AI, others that remain deeply human. The index gives you a framework for identifying which is which in your specific role, rather than treating “AI” as a monolithic force affecting all jobs equally.
Where MindStudio Fits for Teams Acting on This Data
Understanding your exposure score is useful. Doing something with that information is what actually matters.
For teams that have looked at the Anthropic AI Exposure Index and identified specific workflows where AI can augment their work, MindStudio is a practical place to start building. It’s a no-code platform that lets you create AI agents connected to real business systems — without writing code or managing API integrations.
If the index tells you that document drafting, research synthesis, or data analysis are high-exposure tasks in your team’s workflow, MindStudio lets you build a specific agent to handle exactly that process. You can connect it to your existing tools — Salesforce, Google Workspace, Notion, Slack — and have it running in an afternoon.
For example, a legal or compliance team that sees high exposure in document review could build an agent that ingests contracts, summarizes key clauses, and flags non-standard terms. A financial analyst team could automate the generation of recurring reports. A content team could build a drafting assistant that follows their specific style guidelines and pulls from approved sources.
MindStudio supports over 200 AI models, including Claude — so teams that specifically want to use Anthropic’s models as the engine for their agents can do that directly. You can try it free at mindstudio.ai.
The point isn’t to automate jobs. It’s to act on what the exposure data is actually showing: which tasks are already being done better with AI, and building reliable workflows around those tasks rather than leaving it to individual improvisation.
Limitations of the Index Worth Knowing
The Anthropic Economic Index is rigorous, but it has real limitations that are worth understanding before drawing hard conclusions.
It Reflects Claude Usage, Not All AI
The index is based on Claude conversation data. It doesn’t capture what’s happening with GPT-4, Gemini, Copilot, or specialized AI tools embedded in specific software products. A software developer might be getting enormous AI assistance through GitHub Copilot that never shows up as a Claude conversation.
This means the exposure scores are likely understated for occupations that use non-Claude AI tools more heavily. Claude is a strong signal, but it’s one signal.
It’s a Snapshot, Not a Forecast
The data reflects current usage patterns. AI capabilities are changing quickly. An occupation that shows low exposure today might look very different in two years as multimodal AI, computer use, and agentic capabilities improve. The index is useful for understanding the present; it shouldn’t be treated as a stable long-term prediction.
Task Exposure ≠ Job Exposure
This is worth repeating. An occupation where 40% of tasks are AI-exposed is not an occupation where 40% of workers will be displaced. Tasks aren’t jobs. Workers adapt, roles evolve, new tasks emerge. The index measures where AI is active today, not the net employment effect.
Frequently Asked Questions
What is the Anthropic Economic Index?
The Anthropic Economic Index is research published by Anthropic in 2025 that analyzes how Claude is actually being used across different job categories. It maps real usage patterns against O*NET’s occupational database to generate exposure scores — estimates of how much of each occupation’s tasks overlap with what AI is currently being used to help with.
How is it different from other AI job impact studies?
Most AI impact studies use expert judgment to assess whether AI could theoretically perform a given task. The Anthropic Economic Index is based on observed usage data — tasks that people are already delegating to or getting help with from Claude. That makes it a present-tense analysis rather than a theoretical prediction.
Does a high AI exposure score mean my job will be automated?
Not necessarily. The index distinguishes between augmentation (AI helping humans do tasks) and automation (AI replacing humans on tasks). Current Claude usage skews heavily toward augmentation. A high exposure score often means your job’s tasks are well-suited for AI assistance — which is an opportunity, not just a threat.
Which jobs show the highest exposure in the index?
Occupations in computer and mathematical fields, legal analysis, financial analysis, and writing-intensive roles show the highest exposure scores. These jobs involve tasks — drafting, analysis, research, coding — that align closely with what large language models like Claude do well.
Can I look up my specific occupation in the index?
You can find your ONET occupational code using the Department of Labor’s ONET OnLine database and cross-reference it with the occupational groups covered in the Anthropic Economic Index. The index provides granular enough findings to locate your broad occupational category and understand the exposure pattern for roles like yours.
What should I do if my job is highly exposed?
Start at the task level. Identify which specific activities in your job overlap most with what AI handles well, and experiment with AI tools on those tasks. Workers who integrate AI into high-exposure tasks tend to work faster and produce more. The risk isn’t AI replacing you — it’s remaining slower than colleagues or competitors who are already using these tools effectively.
Key Takeaways
- The Anthropic AI Exposure Index uses actual Claude usage data, not theoretical assessments, to measure which occupations are most affected by AI today.
- Exposure is measured at the task level — most jobs contain a mix of high-exposure and low-exposure tasks.
- Higher-earning, knowledge-work occupations tend to show more exposure than physical or service-based roles.
- Current usage data shows AI is primarily augmenting workers, not replacing them — but that distinction is worth monitoring closely as capabilities improve.
- Understanding your exposure score is a starting point. The next step is building reliable AI workflows around the tasks where AI assistance is already proven to help.
If you’re ready to move from insight to implementation, MindStudio lets you build AI agents around the specific workflows the exposure data flags — no code required, and free to start.