An analysis of AI exposure scores across 342 BLS-tracked occupations covering 143 million US workers reveals that one-third of the American workforce faces fundamental job transformation. The businesses that orchestrate AI into their operations will absorb the displaced productivity. The rest will become the displacement.
This analysis is built on the Bureau of Labor Statistics Occupational Outlook Handbook — the US government's authoritative survey of the American labor market. The dataset covers 342 distinct occupations representing 143,066,500 workers, with structured data on median pay, education requirements, projected 10-year employment change, and total employment.
Each occupation was scored on a 0–10 Digital AI Exposure scale by an LLM evaluating the BLS's own detailed job descriptions. The scoring rubric measures how much current AI capabilities — primarily digital, primarily software — will reshape each occupation's daily work. A score of 9 doesn't mean the job disappears. It means the job as currently performed ceases to exist.
34.3% of the US workforce — 49 million workers — holds jobs scoring 7 or higher on AI exposure. These are not future projections. These are jobs where AI is actively reshaping the work right now. The question is not whether these roles change. The question is whether the businesses employing these workers are capturing the productivity gain or bleeding it.
AI exposure is not evenly distributed. It falls overwhelmingly on knowledge workers — the people who generate revenue, manage operations, handle finance, and run sales. The people businesses pay the most.
69.3M workers across 163 occupations. Average AI exposure: 6.9/10. 67.7% of these workers — 46.9 million people — are in jobs scoring 7+.
49.7M workers across 81 occupations. Average AI exposure: 2.9/10. Physical work, manual skill, and real-time human presence create a natural barrier to AI disruption.
This is the asymmetry that every business owner needs to understand. The trades — electricians (2), plumbers (2), HVAC techs (2), carpenters (2) — are largely insulated. Their hands do things AI cannot. But the office? The back office that runs finance, sales, marketing, HR, customer service, and operations? That's the blast zone.
When we group the 342 occupations by business function — the way an actual company is structured — the exposure pattern becomes a roadmap for where AI will hit hardest and where orchestration creates the most leverage.
The highest-exposed functional group in the entire economy. Secretaries (8), office clerks (9), bookkeepers (9), financial clerks (9), information clerks (7), and collections (9). These roles are almost entirely digital — scheduling, data entry, document formatting, invoice processing, payment tracking — and every one of these tasks is within current AI capability.
| Occupation | AI | Workers | BLS Outlook |
|---|---|---|---|
| Secretaries & admin assistants | 8 | 3,453,100 | 0% (flat) |
| General office clerks | 9 | 2,646,000 | -7% |
| Bookkeeping & auditing clerks | 9 | 1,613,400 | -6% |
| Financial clerks | 9 | 1,193,000 | -5% |
| Information clerks | 7 | 1,336,600 | -3% |
| Bill & account collectors | 9 | 166,900 | -10% |
Every single occupation in office administration is projected to decline or flatline by BLS. These aren't AI predictions — these are government labor economists looking at the same data. The BLS sees it. The workers don't. The businesses that employ them definitely don't.
Accountants (8), financial managers (7), financial analysts (9), budget analysts (8), cost estimators (8), underwriters (9), loan officers (8), tax agents (8). Finance is fundamentally data analysis, regulatory compliance, and reporting — all areas where AI already operates at or above junior-analyst level.
Customer service reps (9), claims adjusters (7), travel agents (9), bank tellers (7). The core function — answering questions, processing orders, resolving complaints — is fundamentally digital. LLMs already handle complex natural language interactions. BLS projects declining employment across the board.
Marketing managers (8), market research analysts (9), PR specialists (8), advertising sales (7), wholesale reps (7). The data processing, campaign planning, content drafting, and consumer analysis functions are all highly automatable. But strategic decision-making and relationship management provide a buffer — making this a reshape story, not a replacement story.
Executives (6), management analysts (7), project managers (7), HR managers (7), HR specialists (7), training specialists (7), compensation analysts (8). The pattern: the analytical and administrative components of management are highly exposed; the leadership, negotiation, and interpersonal components are not. AI will make fewer managers capable of running more.
Electricians (2), plumbers (2), HVAC (2), carpenters (2), construction laborers (1), maintenance workers (2). Physical work in unpredictable environments. AI cannot crawl into a ceiling, sweat a pipe joint, or diagnose a compressor by feel. These jobs are growing — BLS projects 4-9% increases across the trades.
Here's the irony: the trades are insulated from AI, but they're drowning in back-office work. A plumbing company's techs score 2/10 on AI exposure. But its office manager scores 8. Its bookkeeper scores 9. Its dispatcher scores 7. The business itself is highly exposed — even when the core trade isn't. This is where orchestration becomes existential.
AI exposure correlates directly with compensation. The more you pay someone, the more AI will reshape what they do.
| Pay Band | Avg AI Exposure | Occupations | Workers |
|---|---|---|---|
| < $40K | 3.6 | 34 | 45,245,500 |
| $40K – $60K | 4.4 | 93 | 41,023,000 |
| $60K – $80K | 5.5 | 101 | 24,600,400 |
| $80K – $100K | 6.3 | 41 | 10,890,100 |
| $100K – $130K | 6.6 | 42 | 12,671,800 |
| $130K+ | 6.6 | 29 | 8,613,800 |
Below $40K, workers average 3.6 exposure — mostly physical jobs. Above $80K, exposure jumps to 6.3–6.6. Your $100K+ employees — the analysts, the managers, the accountants, the marketers — are sitting in the highest-exposure band in the economy.
For a 25-person company spending $150K/year on office admin, bookkeeping, and customer service: those roles average 8.5 exposure. AI doesn't eliminate those people tomorrow. But a competitor who orchestrates AI into those functions will do the same work with 60% fewer hours. Your cost structure becomes uncompetitive. Not in five years. Now.
Higher education correlates with higher AI exposure. The credential premium that drove white-collar career paths for decades is now the target.
| Education Level | Avg AI Exposure | Workers |
|---|---|---|
| Bachelor's degree | 6.9 | 35,990,000 |
| Some college | 6.8 | 3,308,700 |
| Master's degree | 6.2 | 3,085,300 |
| Doctoral / professional | 5.1 | 2,944,400 |
| Associate's degree | 5.0 | 2,763,000 |
| High school diploma | 4.2 | 38,707,800 |
| No formal credential | 2.8 | 28,196,700 |
Bachelor's degree holders — 36 million workers — average 6.9 exposure. No-credential workers average 2.8. The pattern is clear: education trained people for knowledge work, and knowledge work is what AI disrupts.
This isn't an argument against education. It's an argument for augmentation. The educated workforce has the cognitive foundation to work with AI systems. But only if those systems exist. A bachelor's-degree accountant with no AI tooling is more exposed than a high-school-diploma electrician who doesn't need any.
The most dangerous signal in this data is when two indicators converge: the BLS independently projects employment decline, and the AI exposure score is 7+. These occupations are under pressure from both market forces and technological transformation simultaneously.
| Occupation | AI | Workers | BLS Decline | Pay |
|---|---|---|---|---|
| Cashiers | 7 | 3,157,200 | -10% | $31,190 |
| Customer service reps | 9 | 2,814,000 | -5% | $42,830 |
| General office clerks | 9 | 2,646,000 | -7% | $43,630 |
| Bookkeeping clerks | 9 | 1,613,400 | -6% | $49,210 |
| Information clerks | 7 | 1,336,600 | -3% | $43,730 |
| Financial clerks | 9 | 1,193,000 | -5% | $48,650 |
| Material recording clerks | 7 | 1,300,800 | -6% | $46,120 |
| Computer support specialists | 8 | 882,300 | -3% | $61,550 |
| Insurance underwriters | 9 | 127,000 | -3% | $79,880 |
| Computer programmers | 9 | 121,200 | -6% | $98,670 |
Combined, these 10 occupations alone represent 15.2 million workers in jobs that are both shrinking and highly AI-exposed. The BLS didn't need an LLM to see this. Their economists reached the same conclusion from demographic and economic data: these roles are contracting. The AI exposure score explains why.
If your business employs bookkeepers, CSRs, office clerks, or financial clerks, you are paying 2024 wages for roles the labor market is already pricing out. Every month you delay AI integration in these functions, your competitors gain margin advantage. This is not disruption theory. This is BLS data confirming what the technology already demonstrates.
The most important table in this entire analysis. These are occupations that score 7+ on AI exposure and the BLS projects employment growth. AI is transforming these roles. And demand is increasing.
| Occupation | AI | Workers | BLS Growth | Pay |
|---|---|---|---|---|
| Software developers & QA | 9 | 1,895,500 | +15% | $131,450 |
| Management analysts | 7 | 1,075,100 | +9% | $101,190 |
| Project management specialists | 7 | 1,046,300 | +6% | $100,750 |
| Market research analysts | 9 | 941,700 | +7% | $76,950 |
| HR specialists | 7 | 944,300 | +6% | $72,910 |
| Financial managers | 7 | 868,600 | +15% | $161,700 |
| Accountants & auditors | 8 | 1,579,800 | +5% | $81,680 |
| Data scientists | 9 | 245,900 | +34% | $112,590 |
| Financial analysts | 9 | 429,000 | +6% | $101,910 |
| InfoSec analysts | 8 | 182,800 | +29% | $124,910 |
| Marketing managers | 8 | 434,000 | +6% | $159,660 |
| Operations research analysts | 9 | 112,100 | +21% | $91,290 |
This is the mechanism of AI disruption. Software developers score 9/10 on exposure — and demand is growing 15%. Why? Because AI makes each developer more productive, total output rises, the market absorbs more software, and organizations need more developers to orchestrate the expanded capability. The same logic applies to analysts, managers, accountants, and marketers.
High AI exposure + growing demand = the role is being reshaped, not replaced. The workers who adopt AI become dramatically more productive. Their employers capture that productivity as margin, speed, or scale. The businesses that don't adopt AI are now competing against opponents who produce 3–5x more output per employee. This is the selection pressure. This is where orchestration separates the survivors from the extinct.
The BLS organizes occupations into 25 industry categories. When sorted by average AI exposure, the map of disruption becomes a clear gradient from pure knowledge work to pure physical work.
| Category | Avg Exposure | Total Workers |
|---|---|---|
| Math | 8.8 | 426,200 |
| Computer & IT | 8.5 | 4,513,700 |
| Legal | 8.0 | 1,312,600 |
| Media & Communication | 7.8 | 1,164,800 |
| Office & Admin Support | 7.7 | 16,488,600 |
| Business & Financial | 7.5 | 9,819,000 |
| Arts & Design | 7.1 | 728,500 |
| Sales | 7.0 | 10,828,000 |
| Architecture & Engineering | 6.3 | 2,406,700 |
| Life & Physical Science | 6.2 | 1,143,800 |
| Management | 6.1 | 12,027,300 |
| Education | 5.9 | 8,295,000 |
| Community & Social Service | 5.0 | 2,519,200 |
| Entertainment & Sports | 4.6 | 803,000 |
| Protective Service | 4.2 | 2,911,800 |
| Healthcare | 4.1 | 17,533,300 |
| Transportation | 4.1 | 13,050,600 |
| Production | 3.6 | 5,706,000 |
| Personal Care | 3.0 | 3,399,100 |
| Maintenance & Repair | 2.7 | 5,008,200 |
| Food Service | 2.7 | 12,022,100 |
| Farming & Forestry | 2.2 | 889,600 |
| Construction | 2.0 | 6,222,900 |
| Building & Grounds | 1.3 | 3,846,500 |
The top 8 categories — Math through Sales — average 7.5 exposure across 45.3 million workers. These aren't obscure niches. These are the functional categories that comprise the operating backbone of every business in America.
The data tells one story with two endings.
A business employs 20 people. Five are in admin/finance roles averaging 8.5 AI exposure. Three are in sales/marketing roles averaging 7.1. The remaining twelve are in field/trade roles averaging 1.8. The business runs on the same tools it ran on in 2023: QuickBooks on one screen, a CRM on another, a spreadsheet for scheduling, email for customer communication, a shared drive for documents.
The competitor across town has the same 20-person structure. But they've orchestrated their tools. Their bookkeeping flows automatically from invoicing through reconciliation. Their customer inquiries are triaged by AI before a human touches them. Their marketing analyst — one person, not three — produces 4x the campaign output because research, drafting, and reporting are AI-augmented. Their office manager handles what used to require an office manager, an admin assistant, and a part-time bookkeeper.
The first business didn't fire anyone. Nobody came and took their jobs. They just slowly became unable to compete on price, speed, or quality. By the time they noticed, the market had already moved.
The same 20-person business, but they connected their tools into a unified intelligence layer. Not a chat interface. Not a single AI feature bolted onto one app. A system that:
The five admin/finance workers still work there. They just produce the output of eight. The three sales/marketing workers produce the output of five. The owner sees a unified dashboard instead of toggling between 12 apps. Nobody got replaced. The business got multiplied.
This is not a technology question. This is a survival question. The BLS data shows employment declining in high-exposure office roles. The AI exposure scores show why. The growth-despite-exposure data shows that orchestrated workers produce more and are in higher demand. The businesses that orchestrate AI into their operations will absorb the productivity gain, reduce costs, and scale. The businesses that don't will pay 2024 wages for work that the market is repricing to zero.
Liaison is not an AI feature. It is not a chatbot. It is not another SaaS tool. Liaison is the orchestration layer that determines which ending your business gets.
68% of businesses say they "use AI." 7% have it wired into operations. The gap between those numbers is the extinction zone.
The 61% in between have a ChatGPT tab open. Maybe they use an AI writing tool. Maybe their CRM added an AI feature. But their bookkeeper still manually reconciles invoices. Their office manager still toggles between 12 apps. Their sales team still manually updates CRM records. Their marketing team still spends 60% of its time on data processing instead of strategy.
These businesses have AI adjacent to their operations. They need AI inside their operations.
Liaison connects to the tools businesses already use — QuickBooks, HubSpot, Jobber, Stripe, Google Workspace, Slack, and 3,100 others. No rip-and-replace. No $50K implementation. No 6-month migration. Plug into what you have.
All data from all tools is normalized into one queryable intelligence layer. Revenue from QuickBooks, customers from CRM, jobs from scheduling, communications from email — unified into a single operational picture. Ask "what's my cash position?" and get an answer in seconds instead of 2 hours of spreadsheet work.
Liaison monitors operations continuously. Revenue decline, overdue invoices, low utilization, invoicing backlog, customer churn signals, pipeline stalls — detected automatically and routed to the right person with context and recommended action. Not after you ask. Before you need to.
Every business on Liaison makes the system smarter. Cross-tenant pattern recognition identifies what works across similar businesses while keeping individual data private. Your AI doesn't just learn from you — it learns from every business like you.
Automation that goes beyond "if this, then that." Role-based AI agents that understand your business — owner, CFO, ops manager, sales rep — handling the analytical and administrative work that scores 7+ on AI exposure so your people focus on judgement, relationships, and the physical work AI cannot do.
Every high-exposure business function maps directly to a Liaison capability:
| Business Function | Exposure | What AI Replaces | What Liaison Orchestrates |
|---|---|---|---|
| Office & Admin | 8.5 | Scheduling, data entry, document management, correspondence | Unified dashboard, automated workflows, AI-drafted communications, document processing |
| Finance | 8.1 | Invoice processing, reconciliation, compliance reporting, expense tracking | Real-time cash position, anomaly detection, automated reconciliation, cash flow forecasting |
| Customer Service | 8.0 | Inquiry handling, order processing, complaint resolution, FAQ responses | AI triage, contextual routing, sentiment detection, automated resolution for routine cases |
| Sales & Marketing | 7.1 | Research, content drafting, campaign analysis, prospect profiling, reporting | SDR agents, marketing orchestration, pipeline intelligence, cross-channel analytics |
| Management | 7.0 | Data aggregation, KPI tracking, project status updates, HR administration | Executive dashboard, team performance analytics, automated status reports, utilization tracking |
| Field Services | 1.8 | Almost nothing — the hands stay human | Back-office orchestration so the business around the trade runs on AI even though the trade itself doesn't |
Liaison doesn't automate the 1.8-exposure trade work. It orchestrates the 7–9-exposure office work that surrounds the trade work. A plumber's hands are safe. But the business that dispatches, invoices, markets, and manages the plumber is 80% knowledge work — and that knowledge work is in the blast zone. Liaison is the blast shield.
The Bureau of Labor Statistics doesn't have an opinion about AI. They have data. And their data says the same thing the AI exposure scores say: office administration is contracting, financial clerks are contracting, customer service is contracting, bookkeeping is contracting. These aren't predictions. These are measurements of trends already in motion.
The businesses that survive this transition will be the ones that stopped treating AI as a feature to evaluate and started treating it as infrastructure to deploy. Not "should we use AI?" but "how fast can we orchestrate it into every knowledge-work function we operate?"
The data is unambiguous:
Liaison exists to close that gap. Not with another chat interface. Not with another point solution. With the orchestration layer that connects your tools, normalizes your data, watches your operations, learns from your industry, and acts before problems become crises.
Every business in America is making this decision right now, whether they realize it or not. The ones who orchestrate AI into their operations will produce more with less, move faster, see further, and compound their advantage every month. The ones who don't will watch their margins erode, their response times lag, and their market share transfer — gradually, then suddenly — to the businesses that made the other choice.
Orchestrate or go extinct. The data has spoken.
Occupational data: Bureau of Labor Statistics, Occupational Outlook Handbook (2024–2034 projections). 342 occupations, structured fields including median pay, education requirements, employment count, and 10-year projected change.
AI Exposure scoring: Each occupation's full BLS description was evaluated by Gemini Flash (via OpenRouter) using a structured rubric scoring digital AI exposure on a 0–10 scale. The rubric calibrates against anchor occupations (roofer = 1, software developer = 9, data entry clerk = 10) and evaluates the proportion of work that is fundamentally digital versus physical. Original pipeline by Andrej Karpathy (github.com/karpathy/jobs).
Categorization: BLS's 25 standard industry categories were used for industry-level analysis. Business function groupings (Office & Admin, Finance, etc.) were constructed by mapping BLS occupations to operational roles within a typical business.
Limitations: AI exposure scores are LLM estimates, not rigorous economic predictions. They do not account for demand elasticity, regulatory barriers, social preferences, or the rate of technology adoption. The scores measure potential exposure, not guaranteed impact. "High exposure" means the work will be reshaped — not that the job will disappear.
[1] Bureau of Labor Statistics. Occupational Outlook Handbook. U.S. Department of Labor, 2024–2034 projections. bls.gov/ooh/
[2] Karpathy, A. US Job Market Visualizer. GitHub, 2025. github.com/karpathy/jobs
[3] Eloundou, T., Manning, S., Mishkin, P., & Rock, D. "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models." arXiv:2303.10130, 2023.
[4] Deloitte. "AI Orchestration Market Analysis." Deloitte Insights, 2026. AI orchestration market estimated at $8.5B in 2026, projected $35B by 2030.
[5] McKinsey Global Institute. "The State of AI in Early 2024." McKinsey & Company, 2024. 68% of organizations report using AI; 7% report full operational integration.
[6] Felten, E., Raj, M., & Seamans, R. "Occupational, Industry, and Geographic Exposure to Artificial Intelligence." NBER Working Paper 31314, 2023.