Understanding work as a computational system.

Human work is rarely clean. It unfolds across conversations, documents, software, habits, interruptions, decisions, and institutional memory. Most software captures only fragments of that process.

Primary question

How can computers develop an accurate, evolving, and useful understanding of how people and organizations actually work?

Working hypotheses

• Work can be represented as a changing network of events, artifacts, people, decisions, and commitments.
• Long-term operational memory is more valuable than isolated prompts or short-lived context windows.
• AI should observe and learn a workflow before attempting to automate it.
• Trust requires provenance, permission boundaries, and reversible actions.
• Systems should adapt to human work rather than requiring people to reshape their work around software.

Evaluation

Research is grounded through practical evidence: whether a system preserved knowledge, reduced repeated work, improved a decision, surfaced an overlooked risk, or helped a person act with greater confidence.