AI Trends on the World’s Work Marketplace: How AI Is Reshaping the Way Humans Work

AI isn’t replacing humans; it’s reshaping the way humans work. That’s the core finding of our latest research at the Upwork Research Institute. Amid constant headlines predicting mass automation and AI taking away jobs, our platform data tells a more nuanced — and more actionable — story: AI is augmenting work, creating net-new opportunities for people, and catalyzing both professionals and businesses to evolve.

Drawing from millions of jobs and billions in freelancer earnings across more than 130 categories of work, our analysis reveals several key trends:

1. The combination of humans and AI earns trust and wins work.
2. AI is fueling growth in both technical and non-technical fields.
3. AI substitution is real, but limited and transitional.
4. Job evolution is creating new generalist opportunities for professionals.

With this report, we’re building upon our prior research, which demonstrated that AI is not just reshaping technical fields but is also quickly expanding into non-technical domains like design, marketing, and translation. In this new research, we drilled down further to understand how AI has been changing the nature of knowledge work across our platform over the past six months¹.

Observing client and freelancer behavior in this robust dataset gives us many indicators of what’s at play in the broader labor market, and the thriving freelance workforce illuminates the skills, hiring patterns, and work opportunities arising in the new age of AI.

Human+AI earns trust and wins work 

While interest in how much AI can take on grows, there’s no question humans still need to be involved in delivering work. Our clients express greater trust in outcomes where humans remain in the loop, and the trust gap between AI alone and AI+humans is widening (see Figure 1). Clients place more than double the amount of trust in human-AI work than they do in AI-only outputs. At the same time, acknowledgement of the productivity and quality gains AI-enabled workers can offer is growing. We see trust in human-AI collaboration now rivals the trust placed in human-only work. Our internal research also shows that 66% of clients express high levels of trust in outputs from freelancers using AI tools to deliver work, with only 26% trusting work delivered by AI alone².

‍

Figure 1: Upwork clients’ response on a 7-point scale (with 1=not at all and 7=very much) to the question: "How much would you trust the work delivered by..."

External research further supports this. Per a global KPMG study, the majority of people (54%) simply do not trust AI systems today. This is in part driven by the fact that top-tier large language models (LLMs) still currently experience average error rates of 10-15%. As many experts and academics have argued, continuing to have humans in the loop over the long term is critical to preserving trust and other uniquely human dimensions of work.

The research is clear: The real edge lies in human-AI collaboration, where trust and results outperform those achieved by AI alone. Our findings indicate that while AI is poised to take on more tasks, clients continue to rely on human expertise — especially in fields like corporate law, financial planning, design, and project management — to lead, validate, and deliver outcomes.

AI is fueling growth in both technical and non-technical fields

Professionals using AI to enhance their work is becoming increasingly common across numerous fields. When people use AI tools to support their work, the results are better, and clients tend to trust that work more.

Upwork customer insights also illustrate a shift toward augmentation over automation. We found freelancers are using AI for work augmentation 71% of the time, and automation only 29% of the time. The methodology behind this finding aligns with Anthropic’s Economic Index that assesses conversations with Claude, which also shows that the majority of AI use leans toward augmentation compared to automation.

Because of the augmenting effect of AI, combined with upskilling in uniquely human or “power” skills, we’re seeing over 5% year-over-year growth across many non-technical categories on our platform, a significant portion of which were assumed to be ripe for disruption, automation and replacement due to AI:

• In both Virtual Admin Assistance and Project Management, people don’t trust AI to fully manage their work and personal information. However, AI tools are making assistants and project managers more productive.
• Graphic & Presentation Design and Product Design are being augmented, as human creativity is boosted by AI-powered design tools.
• In complex Corporate & Contract Law, sensitivity and nuance require a human touch and critical thought, though AI tools are helping legal professionals complete rote tasks with fewer errors.
• Financial Planning and Accounting & Bookkeeping represent high-stakes work where human judgment, advice and decision-making are crucial. But, those finance and accounting professionals are becoming more effective, analytical and accurate by applying AI tools.

‍

Figure 2: Year-over-year trends in freelancer earnings within categories of work. Darker shades of blue denote higher growth.

We’re also seeing significant growth in more technical fields within AI & Machine Learning and AI Integration. In the first quarter of 2025, Gross Services Volume (GSV) in AI-related work on Upwork grew 25% year-over-year, while growing 52% year-over-year in the Prompt Engineering subcategory. Upward trends are accelerating in generative AI modeling, AI agent design, supervised learning, and multimodal AI development, with no end in sight to opportunities in these burgeoning fields.

Activity on our marketplace also shows that companies are specifically looking for AI builders. Two of the top three platform AI search queries over the last month are “AI agent” and “AI developer.” These searches are met with a deep talent pool of approximately 80,000 global AI specialists on Upwork who have expertise in designing, developing, deploying, and training AI systems.

What’s more, freelance professionals doing AI-related work on Upwork continue to command higher earnings. On average, they receive more than a 40% rate premium per hour compared to freelancers doing non-AI-related work — another overwhelmingly positive augmentation effect from AI.

AI substitution is real, but limited and transitional

With any technological transformation comes natural attrition of rudimentary tasks that can be automated. This progression causes professionals in affected fields to adjust and diversify their skill sets, redefining the human role in impacted work categories. This causes temporary instability in demand within a select few categories, where our platform data indicates that the lowest-value, most transactional jobs are being — and will continue to be — substituted.

Our previous research revealed AI substitution in lower-value, more rudimentary work found in both the Writing and Translation categories. When looking at the last six months, this trend continued, specifically in Content Writing as well as Sales & Marketing Copywriting. This is due to the rapid advances by generative AI models like ChatGPT and others in writing quality for professional settings.

Broader industry research has long predicted this level of automation, so this disruption isn’t surprising. The introduction of more task-specific tools has accelerated this trend in other categories. For example, we saw categories like Market Research and Data Extraction affected by the February 2025 launch of ChatGPT’s Deep Research capabilities. We observed similar trends in Photography, due to image generation capabilities in tools like DALL-E, Adobe Firefly, and Canva becoming viable options for business purposes.

‍

Figure 3: Year-over-year trends in freelancer earnings within categories of work. Darker shades of orange denote higher levels of replacement or substitution effects.

Categories where substitution of lower-complexity, lower-value work is occurring only account for a very small percentage of Upwork’s total GSV. As AI substitutes for lower-level tasks in those categories, freelancers are embracing the tools and upskilling necessary to move into higher-level work such as designing complex research, data analysis, content strategy, or creative art direction. These jobs rely less on the nuts and bolts of execution (more ably handled by specific AI tools) and more on complex, uniquely human skills like strategic thinking, communication, and creativity.

This opens up the aperture for workers in fields experiencing AI substitution, giving them opportunities to grow professionally. In fact, our previous platform findings highlighted 268% year-over-year growth in AI-related jobs in non-technical fields like design, marketing, and translation.

Job evolution is creating new generalist opportunities

Some categories of work that aren’t impacted by immediate augmentation or substitution from AI are still experiencing a correlated evolution of jobs and skills. For example, we’re seeing this take shape in Web, Mobile & Software Development through changing demand for certain services or skill sets.

Repetitive code generation work is less sought-after due to LLMs and agents like Anthropic’s Claude and OpenAI’s Codex, but clients are still looking for experienced coders for higher-complexity and generalized integration needs. In fact, we found that freelancers with jobs requiring coding skills for at least 25% of the work now earn 11% more for the same jobs compared to November 2022, when ChatGPT launched. This is evidence that clients are having freelancers put coding skills to higher-order, broader use, rather than on routine, lower-value coding tasks.

Other evolving trends and skills that are migrating work within the category include vibe coding (the modern-day low-code/no-code approach) using tools like Mistral Code, Lovable, and Bolt. Traditional coding skills aren’t necessarily required for the work, and the work opportunities shift to a different subcategory. While this might initially present as substitution in the original coding-based subcategory, it’s offset by augmentation or net-new work cropping up in a tangential non-coding-based subcategory like data analysis and testing, IT security and compliance, or AI app integration.

Work category evolution like this leads to the rise of the “generalist,” where clients are looking less for the traditional software engineer and more for talent who can work with AI to code, design, and apply uniquely human skills like creativity, business logic, critical thinking, communication, and problem-solving. Once again, we’re seeing that while AI is influencing the job to be done within these evolving categories, the work still requires human involvement that can’t be replicated by AI.

Adaptation for the human + AI workforce

This research brings clarity on what’s really going on with AI and work. Signals from the Upwork marketplace tell us that AI’s impact on work is omnidirectional, varying depending on the category of work and the skills required. The augmentation-substitution-evolution framework visible on Upwork today can help leaders and professionals understand new nuances in hiring, skilling, and staying ahead in an increasingly AI-enabled market.

To further accelerate the study of AI’s impact on work and labor markets, we’ve established an Upwork Economic Advisory Council that brings together leading academics to advise on human-AI workforce design, identify labor market shifts, and guide advancement of Upwork’s human and AI-powered platform. To learn more about the Economic Advisory Council and all of the Upwork Research Institute’s latest research on work trends, visit upwork.com/research.

About the author

Kelly Monahan 

Dr. Kelly Monahan is the Founder and Managing Director of the Upwork Research Institute, where she leads research on emerging technologies, workforce transformation, and the rise of independent professionals. Previously, she held leadership roles at Meta, Accenture, and Deloitte, shaping future of work strategies for global organizations. She’s the author of two books, including the USA Today bestseller Essential, and a sought-after keynote speaker on talent innovation. Kelly holds a Ph.D. in organizational leadership and blends behavioral science with workforce strategy.

‍

¹ Upwork platform data is organized across three levels. Level 1 (L1) categories represent broad domains of work, Level 2 (L2) captures specialized areas within those domains, and Level 3 (L3) reflects task-level attributes. 62 Level 2 categories represent a robust sampling of work taking place in knowledge settings. As a reference point, the Bureau of Labor Statistics groups major occupations into 23 categories. Our analysis enables us to look deeper than high-level occupation classifications to show more specifically how roles, skills, and tasks are evolving alongside AI.

² Client survey conducted in October 2024 asking the degree of trust clients have with freelancer-delivered and AI-delivered work. (n=673)