Scaling Global Innovation Centers for Future Growth

The COVID-19 pandemic and accompanying policy steps caused economic disruption so plain that sophisticated statistical techniques were unnecessary for numerous concerns. For example, joblessness jumped sharply in the early weeks of the pandemic, leaving little room for alternative explanations. The effects of AI, however, may be less like COVID and more like the internet or trade with China.

One common approach is to compare outcomes between more or less AI-exposed employees, firms, or industries, in order to separate the impact of AI from confounding forces. 2 Direct exposure is normally specified at the task level: AI can grade research however not manage a classroom, for example, so teachers are considered less reviewed than workers whose entire task can be carried out from another location.

3 Our technique integrates data from three sources. Task-level direct exposure estimates from Eloundou et al. (2023 ), which measure whether it is in theory possible for an LLM to make a task at least twice as fast.

Leveraging AI to Improve Market Forecasting

4Why might real usage fall short of theoretical ability? Some tasks that are in theory possible may disappoint up in use since of design limitations. Others might be slow to diffuse due to legal restraints, particular software requirements, human confirmation steps, or other obstacles. Eloundou et al. mark "License drug refills and provide prescription info to pharmacies" as fully exposed (=1).

As Figure 1 programs, 97% of the tasks observed throughout the previous four Economic Index reports fall into classifications rated as in theory possible by Eloundou et al. (=0.5 or =1.0). This figure shows Claude use distributed throughout O * NET tasks organized by their theoretical AI direct exposure. Jobs ranked =1 (totally feasible for an LLM alone) represent 68% of observed Claude usage, while tasks rated =0 (not feasible) account for simply 3%.

Our new step, observed exposure, is indicated to measure: of those jobs that LLMs could in theory accelerate, which are really seeing automated use in expert settings? Theoretical capability incorporates a much broader series of tasks. By tracking how that gap narrows, observed exposure offers insight into economic changes as they emerge.

A task's exposure is higher if: Its tasks are theoretically possible with AIIts jobs see significant use in the Anthropic Economic Index5Its tasks are carried out in job-related contextsIt has a fairly higher share of automated use patterns or API implementationIts AI-impacted tasks comprise a bigger share of the overall role6We provide mathematical information in the Appendix.

Vital Expansion Statistics to Track in 2026

We then adjust for how the task is being performed: totally automated executions get complete weight, while augmentative usage gets half weight. Lastly, the task-level protection steps are balanced to the profession level weighted by the fraction of time spent on each job. Figure 2 shows observed exposure (in red) compared to from Eloundou et al.

We compute this by first averaging to the occupation level weighting by our time fraction step, then averaging to the profession classification weighting by total employment. The measure reveals scope for LLM penetration in the bulk of tasks in Computer system & Math (94%) and Office & Admin (90%) professions.

Claude currently covers just 33% of all tasks in the Computer & Math category. There is a big exposed location too; lots of tasks, of course, remain beyond AI's reachfrom physical agricultural work like pruning trees and operating farm machinery to legal jobs like representing customers in court.

In line with other data revealing that Claude is extensively utilized for coding, Computer system Programmers are at the top, with 75% coverage, followed by Client service Representatives, whose primary jobs we increasingly see in first-party API traffic. Data Entry Keyers, whose primary task of reading source files and getting in information sees significant automation, are 67% covered.

How to Analyze the Global Market Landscape

At the bottom end, 30% of workers have no coverage, as their jobs appeared too rarely in our information to satisfy the minimum limit. This group includes, for instance, Cooks, Motorcycle Mechanics, Lifeguards, Bartenders, Dishwashers, and Dressing Space Attendants. The US Bureau of Labor Stats (BLS) releases routine work projections, with the most recent set, released in 2025, covering predicted changes in employment for every occupation from 2024 to 2034.

A regression at the occupation level weighted by current work finds that development projections are somewhat weaker for tasks with more observed direct exposure. For every single 10 percentage point boost in protection, the BLS's growth forecast stop by 0.6 portion points. This provides some recognition in that our measures track the individually obtained quotes from labor market analysts, although the relationship is slight.

Each strong dot reveals the typical observed direct exposure and projected work modification for one of the bins. The dashed line shows a simple linear regression fit, weighted by current work levels. Figure 5 shows qualities of employees in the leading quartile of direct exposure and the 30% of employees with zero exposure in the three months before ChatGPT was released, August to October 2022, utilizing data from the Existing Population Study.

The more unveiled group is 16 percentage points most likely to be female, 11 percentage points more likely to be white, and almost twice as likely to be Asian. They make 47% more, typically, and have higher levels of education. People with graduate degrees are 4.5% of the unexposed group, however 17.4% of the most unveiled group, a practically fourfold difference.

Scientists have taken different methods. Gimbel et al. (2025) track changes in the occupational mix utilizing the Present Population Survey. Their argument is that any essential restructuring of the economy from AI would appear as changes in circulation of jobs. (They discover that, up until now, modifications have been average.) Brynjolfsson et al.

Mapping Future Shifts of Enterprise Trade

( 2022) and Hampole et al. (2025) use task posting data from Burning Glass (now Lightcast) and Revelio, respectively. We concentrate on unemployment as our concern result due to the fact that it most directly records the capacity for financial harma worker who is unemployed desires a task and has not yet found one. In this case, job postings and employment do not always signify the need for policy actions; a decrease in job postings for an extremely exposed role may be neutralized by increased openings in an associated one.