Help Build the First
Real‑World Benchmark
for White‑Collar AI
Shape the future of AI in business practice.
AI systems ace tidy lab tests yet still struggle with the messy, contextual work you tackle every day—pricing a deal, structuring a partnership, forecasting sales from imperfect data. By donating anonymized artifacts from genuine business tasks, you help create BizBench, the first benchmark that measures how well AI agents perform real multi‑step knowledge‑work workflows.
Your contributions close the "disciplinary gap" between AI research and business reality that experts highlight as the next frontier for progress. Only practitioners like you understand the nuanced, contextual challenges that make real business work fundamentally different from academic test cases.
Real-World Context
The messy, imperfect data and constraints that define actual business decisions
Domain Expertise
Deep understanding of workflows, priorities, and trade-offs in your field
Interdisciplinary Bridge
Connecting rigorous research with practical business applications
Recognition & Impact
Join the select group of business professionals advancing AI science. We will gratefully acknowledge all qualified contributors (with your consent) in the final research article and public releases of BizBench—recognizing you as an integral part of the project's success.
1. Create Profile
Share your professional background and Excel expertise to help us understand your business domain knowledge
2. Design Tasks
Create real-world business tasks with evaluation criteria that challenge AI systems in meaningful ways
3. Submit Solutions
Upload your Excel work and provide expert evaluation scores to build the benchmark dataset
Data Stewardship
Every submission is passed through a multi‑step pipeline that strips or hashes names, emails, account numbers, addresses, and any other personally identifiable information. Our team then reviews each artifact to verify that no sensitive details remain before it enters the research corpus.
Research Participation Consent
Please provide consent to participate in this research