Gen AI’s Impact on U.S. Employment and Office Space

The discourse surrounding Gen AI's economic impact is rapidly evolving, drawing upon a growing body of research. Foundational research such as "Occupational Heterogeneity in Exposure to Generative AI" by Felten E., Raj and Seamans considers the characteristics of each occupation and the degree those occupations are exposed to language modeling and image generation. They suggest that highly educated, highly paid, white-collar occupations may be the most exposed, while the disruption would differ across demographic groups, with a higher percent of female, White and Asian employment at risk.

More recently, studies have attributed slow and declining job growth for early-career workers to Gen AI. A study by the Burning Glass Institute¹ reveals that Gen AI may have eroded the value of college degrees as college graduates struggle to secure employment opportunities. The sharpest increases in unemployment rate are concentrated in sectors like tech, business operations and finance, where output has grown strongly, without matching job growth. The turning point was seen during the winter of 2022 following the launch of ChatGPT. Another recent study² identifies a 13% decline in employment for early-career workers aged 22 to 25 in occupations highly exposed to language models such as customer service, accountants and software developers.

The Bureau of Labor Statistics (BLS) released its 2024-2034 employment projections in August 2025, providing a critical lens to view the future of the U.S. labor market. BLS expects that “demand for AI-based systems, data processing, software development, research services and associated consulting services will drive robust employment growth in professional, scientific and technical services (+7.5%) and the information sector (+6.5%).” These two sectors have been leading AI adoption as the information sector drives AI development including model building and system development, while professional, scientific and technical services focus on AI deployment and business applications.

Figure 1: Occupational employment size in 2024 and change from 2024-2035, projected % change

Conversely, the BLS projects slow and declining job growth for government, retail trade and administrative and support roles. This contraction is directly attributed to ongoing technological advancements, particularly the acceleration of automation and the continued expansion of e-commerce. The dual trajectory underscores a crucial aspect of technological advancement: it acts as a catalyst for both growth and decline. Existing demographic shifts require more healthcare services and Gen AI will likely serve that sector’s growth. Meanwhile, government, retail trade and administrative roles have been undergoing automation, and Gen AI will accelerate their decline.

Before we make assumptions about Gen AI, a review of past technological impacts should be helpful. Gen AI is largely recognized as a degree of advancement we have never seen before. However, it is not the first and likely will not be the last of the never-seen-before technologies in history. If the scale and speed of AI disruption is more profound than ever before, it is important to compare it with historical examples to expedite strategic planning and business adaptations.

The introduction of personal computers and spreadsheet software in the 1980s and 1990s revolutionized office work. Tasks that once demanded hours of manual effort could suddenly be completed in a fraction of the time, leading to falling demand for traditional clerical roles. Similarly, the advent of smartphones in the early 2000s ushered in an era of constant connectivity and enabled widespread remote work, further altering established job functions and expectations.

In the present landscape, Gen AI is expected to further automate routine tasks, particularly within office administration, leading to the continued displacement of traditional roles that have already been on a decline due to previous technological advancements. The sophisticated capabilities of Gen AI, such as content creation and data transformation, could potentially render more complex roles traditionally categorized as higher value obsolete.

Nevertheless, this does not necessarily translate to a net decrease in overall employment. Instead, it signals a significant shift in the types of skills and roles that will be in demand like we have experienced many times before. As AI assumes more repetitive and automatable tasks, the demand for positions requiring critical thinking, creative problem-solving and advanced technical skills is anticipated to grow.

To better understand AI’s impact across different job sectors, we analyzed the Annual Social and Economic Supplement (ASEC) data from the Census’ Current Population Survey (CPS), focusing on historical occupational shifts linked to technological advancements. Occupations can be categorized into four distinct groups based on the exposure level to Large Language Models (LLMs) and the ability to integrate with LLM-based AI tools.³ This framework helps to identify which roles will evolve with new responsibilities and which may face redundancy.

Figure 2: Analysis of occupational exposure to LLM and integration potential with AI tools

High exposure to AI and high integration

This category encompasses jobs that are significantly affected by AI and are deeply integrated into AI systems. These roles are expected to undergo a substantial efficiency boost, necessitating upskilling and adaptation to new AI tools.

A prime example is a software developer. Since the occupation entered the survey in 2003, it has only grown with technology. Widespread applications of tablets, smartphones and wearable digital devices created surging demand for software developers. When software developers leave a job, there is a 98% likelihood they will be rehired for the same or a similar position, including roles like computer scientist, system support, web developer or programmer.

Figure 3: Employment of software developers, million persons

The wide adoption of Gen AI likely accelerated and extended demand for software developers. Almost all industries and businesses are looking to create AI solutions and enhance productivity through automation. For jobs with high AI integration potential as software developers, AI functions as a powerful co-pilot, assisting with code generation, debugging, and documentation. This does not eliminate human involvement but rather amplifies human expertise, enabling workers to tackle more complex projects and achieve significantly more output.

The phenomenon has most recently created a bifurcation within the tech industry. While the experienced tech talents have become incredibly popular and sough after, entry-level roles have been cut back. CBRE’s “Scoring Tech Talent 2025” report shows the total number of tech job postings have fallen while the share of AI-related job postings has doubled to more than 20%.⁴ Among these tech job postings, the share of remote roles dropped to 16% from a peak of 23% in mid-2022. In San Francisco, the remote share fell further to 10% as AI-related companies overwhelmingly require full-time in-office work. This has revived office demand in the San Francisco Bay Area. Other office markets with significant tech talent will likely follow.

High exposure to AI and low integration

Jobs in this category are particularly vulnerable to displacement, as AI can readily automate their repetitive tasks. Consider the office and administrative support occupations. These roles have already been experiencing a decline due to prior technological advancements and AI's ability to perform complex administrative tasks will only accelerate this trend. Similarly, computers rendered typists obsolete, reducing their numbers from over a million in the 1970s to almost none. The majority of the displaced typists became secretaries and administrative assistants. However, the occupation of secretaries and administrative assistants has long been challenged and replaced by other technologies, such as the internet and smartphones.

Figure 4: Employment of secretaries and administrative assistants, million persons

Moreover, some previously growing occupations have reversed trends and declined since 2023 with the rising popularity of AI tools, including customer services representatives, paralegals, marketing and sales managers, and computer systems administrators. Gen AI will likely drive further declines in these support, administrative and automatable roles. Workers in these roles have historically stayed in similar occupations while slowly transitioning into miscellaneous management and sales functions. They represent a substantial portion of traditional office-using jobs, and the projected decline will directly reduce the demand for general-purpose office space, particularly for functions centered on routine administration.

Low exposure to AI and high integration

This group consists primarily of service workers who use limited or flexible office space and will likely experience major job growth in the long run due to new services surrounding AI operations and functions. Positions such as IT support, maintenance technicians and taxi drivers fall into this category.

While AI automates, it also necessitates ongoing maintenance, oversight and integration. Figure 5 shows the impact of ride apps on taxi drivers. When Uber launched in 2010, taxi drivers saw small declines at first, but the demand for drivers exploded a few years later. In 2025, there were nearly 650,000 taxi drivers, a 62% increase from 2010. The ride apps transformed the occupation with price and earnings differentiation based on service qualities.

Figure 5: Employment of taxi drivers, thousand persons

Even with the rise of autonomous vehicles, drivers are required to support and safeguard the system. The growth of AI inherently drives an increase in demand for supporting and testing roles. This indicates a continued, and potentially expanding, need for specialized technical talent and, by extension, dedicated office or data center space that supports critical IT infrastructure, cybersecurity and AI system management.

Low exposure to AI and low integration

This group comprises jobs that are neither vulnerable to AI replacement nor heavily integrated into AI systems, likely experiencing the least transformation.

Examples include healthcare and childcare workers and manual labor roles. Ninety-nine percent of physicians and surgeons (Figure 6) stay in the healthcare industry through job changes. The creation of WikiHow, wearable health devices and virtual care services could not disrupt the rising demand for the in-person services provided by healthcare professionals.

Figure 6: Employment of physicians and surgeons, thousand persons

While these roles are less directly affected by Gen AI's core capabilities (text and image generation), AI's broader influence on supply chains, research and consumer behavior could indirectly change the skillset of such roles. For instance, AI-driven design tools could alter workflows for creative professionals or AI in robotics could impact healthcare and industrial production processes.

Overall, workers often stay with the same or similar occupations in the near term, therefore finding a growing and resilient occupation to enter is the surest way to avoid being disrupted out of a job, and actively upskilling with technological advancements is the next step. Historically, the most stable job category has been healthcare practitioners and technical occupations. The least stable and more likely tech-disrupted categories included office clerks, administrative support, installation, maintenance and cleaning workers. Such job losses have largely been absorbed into non-office using roles such as retail sales, trade, food preparation and other service occupations, but many roles will likely continue to experience augmentation as well as substitution from technologies like AI. As AI likely put greater emphasis on critical thinking, adaptability and ethical choices, the emergence and expansion of AI-related jobs, such as prompt engineers and infrastructure architects, are expected to outgrow displacement and disruption.

The interplay between shifts in job functions, increased productivity per employee and evolving work methodologies will dictate the future trajectory of office demand, rent growth and asset valuations.

As discussed earlier, while some roles decline, others are growing, particularly in the tech industry which includes AI development itself. AI fosters the emergence of entirely new office-using roles. The critical factor is the productivity gain enabled by AI. If AI allows a "super-worker" (e.g., a software developer or data analyst) to achieve significantly more output, fewer individuals might be needed to accomplish the same volume of work. This increased efficiency per employee could lead to a reduction in workers and thus space needs or a stabilization of overall office demand despite economic growth, as the same output is achieved with a smaller footprint.

Analysis of AI adoption trends across key office-using sectors provides further evidence of this evolving dynamic. Figure 7 shows the level of business adoption of Gen AI in producing goods and services. Information and professional services sectors lead as the AI economy drives a boost of services demand—this is particularly note-worthy as these businesses are major office tenants. Markets like San Francisco are benefiting greatly from the rise of AI development. Other industries such as arts, entertainment and healthcare are slower to adopt but exhibit upward trends.

Figure 7: AI adoption, percent of businesses using AI to produce goods and services, rolling three-month average

To create forecasts for office leasing fundamentals, we build assumptions on Accenture’s task-oriented research of how much work can be automated versus augmented by occupation.⁵ Under a 10-year scenario, we assumed that 60% of jobs with “higher potential for automation” could be displaced, while jobs with “higher potential for augmentation” would create 0.5 new positions for each existing one due to knock-on effects of productivity-driven growth. Jobs with “lower potential for augmentation or automation” or jobs with “non-language tasks” were held roughly constant.

The outcome of this exercise indicates a mixed effect across office-using employment categories (Figure 8). Administrative support and sales functions show the most significant negative impact on annual employment in the next 10 years, in line with historical average patterns, while architecture, engineering, design and technology roles show a positive net impact to employment growth over the next 10 years due to AI.

Figure 8: AI impact on annual office employment change, 2026-2034, by office-using occupation category

The aggregate impact across all office-using sectors is negative, but far less severe than pure automation scenarios often portray in headlines. Incorporating the employment change projections, while holding all other variables constant, our office rent growth and vacancy forecasting models make the following suggestions:

While the aggregate impacts appear modest, the analysis reveals significant variation across markets based on employment composition. Metro areas with large concentrations of routine back-office functions—including Tampa, Orlando, South Florida and San Antonio—face greater vulnerability to automation-driven job losses, particularly if displaced roles are not replaced by higher-value positions. Conversely, markets anchored by research institutions, technology clusters and innovation-driven industries—such as San Francisco, San Jose, Boston and Seattle—are better positioned to capture AI-driven job creation, given their concentration of STEM occupations and technical talent.

These dynamics translate directly into divergent vacancy impacts over the forecast period. Compared to our base case, office vacancy in the Bay Area, Seattle and Boston declined more sharply for 2025-2035, reflecting more resilient office demand in the specific AI-driven scenario. Meanwhile, South Florida, Orlando, Tampa and San Antonio experienced a slightly higher terminal vacancy in 2035 as automation-related displacement had a stronger impact on demand.

Figure 9: Office vacancies forecast, H2 2025 House View vs. AI scenario

For real estate investors and developers, these dynamics necessitate a strategic re-evaluation of asset selection, market selection and office asset management. In addition to common area amenities, flexible floor plans and sustainability credentials that have become the standard for modern offices, technology infrastructure and other AI adoption needs should enter the business planning of office development and operations.

We are in an early stage of understanding and forecasting Gen AI’s impact on the labor market and the real estate industry. The analysis presented in this paper underscores a nuanced transformation rather than a simple reduction in employment or office demand.

If history is our guide, job displacement and job augmentation are balancing forces. Productivity gains and business growth are two sides of one coin. Occupations with long-term demand growth and the ability to integrate technological advancements will benefit greatly from AI development. Skills such as critical thinking, creativity and adaptability will become even more important in the human-plus-AI world.

Meanwhile, job switching will be gradual and tied to skill levels. The least stable and more likely tech-disrupted occupations likely include office clerks, paralegals and administrative support. The growing and stable jobs likely include and are not limited to healthcare practitioners, AI system developers and scientific researchers.

For office, Gen AI is yet another accelerator of existing trends for flight to quality and space optimization for evolving tenant demand. Future offices will need to be designed as hubs for collaboration, innovation and deep work, equipped with advanced technological infrastructure to support AI-augmented tasks.

---