Return research task list
M-01In ProgressChina / United States / EuropeCross-industry

How can AI agents integrate into real organizational workflows?

Moving an agent from a demo environment into an organizational workflow typically involves three key gates: permissions, approvals, and accountability. This task tracks the real-world deployment process, documenting critical transition points from pilot to scale-up, including failures and rollbacks that occur at these junctures.

Target Case
30row
Submitted
18row
Deadline
2026-12
Completion progress
60%

Research Methodology

The method determines the form of evidence, and the form of evidence determines how far the conclusion can go.

  • Semi-structured interview
  • Process Node Record
  • Pre- and Post-Deployment Comparison

Delivery Requirements

  • 01Organization Flowchart and Agent Access Points
  • 02Checklist: Transition from Pilot to Scale
  • 03Failure and Rollback Case Records

Acceptance Criteria

  • 01Specify the organization name and deployment scope.
  • 02Specify the proportion of manual intervention for each stage.
  • 03Must include at least one deployment that did not meet expectations

Standard of Evidence

Consistent with A-AO research standards.

  • Each case must have at least 2 pieces of evidence, including at least 1 that are either primary sources or authoritative secondary sources.
  • Resales (including materials provided by enterprises) cannot serve as the sole basis for support and may only be used for comparison.
  • Sources must be traceable: values without sources are excluded from the case body.
  • Evidence levels and case verification levels are recorded separately and cannot be converted into one another.

Task Owner

Dr. Chen Sarah · MIT Media Lab

Method Design and Acceptance Criteria for This Task

View Researcher Directory →