Answers are grounded in this paper and cite their sources.
Agentic artificial intelligence is taking supply chain software from recommendation to execution. Planning and procurement agents now execute replenishment orders, reallocate freight, and negotiate with suppliers based on delegated powers, all without simultaneous human approval. Virtually every organization maintains the records of such activities as log files, subject only to IT retention policies until the files are no longer useful for diagnostic purposes and then disposed of. This opinion article contends that such treatment misclassifies what is, in substance, evidence of business transactions. An agent-executed decision that commits expenditure, alters a contractual position or redistributes risk across trading partners creates obligations that auditors, counterparties, regulators and courts may later test, and its trace therefore meets the functional definition of a record. Drawing on records continuum theory, the article follows a single agent decision across the dimensions of creation, capture, organisation and pluralisation and locates the points at which it leaves any recordkeeping regime. It then specifies the attributes of a trustworthy decision record in terms of reliability, authenticity and integrity, and proposes a governance arrangement in which records and information managers hold design authority over decision records alongside IT, procurement, legal and audit functions. The argument is addressed to information professionals in commercial settings who will increasingly be asked to evidence what an agent did, and on whose authority.
No takes yet. Share an insight, caveat, or question.
Tathavadekar et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: