Toward a Human-AI Task Tensor: A Taxonomy for Organizing Work in the Age of Generative AI
Citation. Doshi, A. R., & Moore, A. P. (2026). Toward a human–AI task tensor: a taxonomy for organizing work in the age of generative AI. In Handbook of Artificial Intelligence and Strategy (pp. 124–142). Edward Elgar Publishing. https://doi.org/10.4337/9781035345885.00014
Summary. Doshi and Moore introduce the Human–AI Task Tensor, a conceptual framework for organizing generative-AI-enabled work across eight interdependent task dimensions: task definition, AI contribution, interaction modality, audit requirement, output definition, decision-making authority, AI structure, and human persona. The chapter uses projections of the tensor to derive practical frameworks including an AI Function Matrix, a Task Augmentation/Automation Scale, a Task Audit Matrix, and a Human–AI Task Canvas. Its main contribution is taxonomic and design-oriented: it gives researchers and organizations a vocabulary for comparing tasks and AI configurations before making claims about augmentation, automation, or collaboration.
Key claims
- The Human–AI Task Tensor defines eight dimensions for analyzing AI-enabled tasks: task definition, AI contribution, interaction modality, audit requirement, output definition, decision-making authority, AI structure, and human persona.
- The AI Function Matrix combines AI contribution and output definition to distinguish six AI task functions: production, idea generation, assistance, editing, explanation, and open-ended interaction.
- The Task Augmentation/Automation Scale elaborates decision-making authority into 20 levels of human–AI arrangement, separating degrees of AI augmentation of human decisions from degrees of AI automation of human input.
- The Task Audit Matrix separates process audit and output audit requirements, yielding four task types: open exchange, verifiable application, process exploration, and expert application.
- The Human–AI Task Canvas is proposed as a way to visualize and compare specific tasks across all eight tensor dimensions rather than treating generative AI use cases as isolated examples.
- The chapter argues that task-level mapping can help researchers organize prior studies and identify future research opportunities by making visible which task dimensions differ across generative AI applications.
Evidence & limitations
- Evidence is conceptual synthesis and framework development rather than an empirical test of task outcomes, adoption, labor substitution, or worker experience.
- The chapter is useful for vocabulary, classification, and design diagnosis, but it should not be treated as evidence that any particular tensor-derived work design improves performance, trust, motivation, learning, or labor demand.
- Crossref lists the source as a 2026 Edward Elgar book chapter in Handbook of Artificial Intelligence and Strategy; arXiv metadata also identifies public preprint versions, while this record uses the publisher DOI as canonical provenance.
- The Drive PDF text extracted cleanly during ingest; the file hash is recorded in frontmatter.