Record initial work such as data access, equipment mapping, integration, reference labeling, local evaluation, and user preparation. Include the time of industrial staff as well as external purchases. Research effort already incurred and additional effort needed for the next step answer different questions and should be visible separately.
Continuing work can include data quality checks, hosting, monitoring, review, incident investigation, model evaluation, software maintenance, and periodic adaptation. Identify who would own that work and whether the pilot relied on unusual researcher attention that a continuing arrangement would need to replace.
Opportunity cost matters when the same people or equipment could support other activities. Explain major competing uses where they influence the next decision. Avoid treating resources as free solely because they come from an existing team or budget.