TLDR: A new study by Upwork, the Human+Agent Productivity Index (HAPI), demonstrates that while AI agents often fail to complete professional tasks independently, their success rates dramatically improve by up to 70% when collaborating with human experts. This research, based on over 300 real client projects, highlights the critical role of human intuition and expertise in the future of AI-powered work.
PALO ALTO, Calif. – November 13, 2025 – Upwork Inc. (Nasdaq: UPWK), a leading global work marketplace, has released the initial findings from its groundbreaking Human+Agent Productivity Index (HAPI). This comprehensive study, the first of its kind, offers a data-driven evaluation of how human expertise significantly enhances the performance of AI agents in real-world professional tasks.
The research indicates that artificial intelligence agents, even those powered by the most advanced language models, routinely encounter difficulties in independently completing straightforward professional projects. However, the study also unveils a highly promising path forward: when these AI agents are partnered with human experts, project completion rates experience a substantial surge, increasing by up to 70%.
Andrew Rabinovich, Chief Technology Officer and Head of AI & Machine Learning at Upwork, commented on the findings, stating, “In our Human+Agent Productivity Index, agents struggle to complete simple real-world projects on their own. However, when paired with expert human professionals, project completion rates improve dramatically, supporting our firm belief that the future of work will be defined by humans and AI collaborating to get more work done, with human intuition and domain expertise playing a critical role.”
The HAPI study’s methodology is particularly noteworthy as it deviates from conventional AI agent evaluations that often rely on fixed or synthetic datasets. Instead, Upwork’s research is built upon an analysis of more than 300 actual client projects executed on its platform. These projects were specifically chosen for their simplicity and well-defined nature, offering AI agents a reasonable chance of success. Such low-complexity tasks represent a small fraction (less than 6%) of Upwork’s total gross services volume.
Also Read:
- New Research Highlights Critical Need for AI Content Guardrails in Enterprises
- Salesforce Study Reveals Data Quality is Pivotal for Employee Trust in AI Adoption
This systematic evaluation marks a significant step in understanding the practical application of AI in professional environments. It challenges both the exaggerated claims surrounding fully autonomous AI agents and the widespread anxieties that such technology will imminently displace knowledge workers. The findings strongly suggest that the future of work is not a competition between humans and machines, but rather a powerful synergy where both collaborate to achieve superior outcomes. Upwork plans to present the full HAPI results at an upcoming NeurIPS workshop.
