FideAI
Enterprise

Fide research · Proposal

From Approval to Action

Can organizations reconstruct who authorized an agent’s actions?

A proposal to examine approval boundaries, handoffs, and recovery in simulated enterprise workflows.

Follow or contribute ↓

The problem

What we want to understand.

A workflow can produce the expected output while containing an unauthorized commitment or an incorrect approval. A final result alone does not show whether the right person retained decision authority along the way.

Fide’s approach

What we want to make measurable.

We want to connect approvals, actions, and accountable owners in evidence a reviewer can inspect. The proposed study distinguishes an agent ignoring a clear boundary from an organization leaving the boundary undefined.

Working hypothesis · to be tested

Explicit approval state and verifiable handoff records may reduce unauthorized commitments and improve reconstruction compared with policy prompts and logs alone. They may add delay without helping when decision ownership is unclear.

Proposed first study

A path from question to evidence.

  1. 01

    Choose one synthetic workflow, such as access provisioning, with practitioners. Define roles, approval limits, decision owners, and rollback rules.

  2. 02

    Compare policy prompts and ordinary logs with explicit approval state and linked handoff records; assess an independently enforced permission check separately.

  3. 03

    Include routine cases, stale approvals, conflicting instructions, and interrupted work while matching workload and available information.

  4. 04

    Score simulated actions against a predefined policy. Design a separate reviewer study if testing whether people can reconstruct the workflow, with appropriate review and consent before recruitment.

What we would measure

  • Unauthorized commitments and approval bypass
  • Handoff fidelity and useful task completion
  • Recovery steps after interrupted or incorrect actions
  • Whether records connect each action to valid approval

Intended public outputs

  • A workflow assurance protocol and reusable synthetic cases
  • An evidence checklist linking approvals, actions, and owners
  • If undertaken, findings on control and useful-work tradeoffs

A proposal using simulated transactions and accounts. It does not establish an enterprise assurance service or production readiness. Human-review outcomes would require a separately designed study.

Research record

An open view of the work.

This page is the current study brief. Substantive updates and released materials will remain here as the project develops.

  1. Proposal developed from FID-080

    The brief narrows the call to approval evidence in a first synthetic workflow. A study lead and practitioner input are needed to choose and validate that setting.

Follow the research

Stay close to the question.

Interested in From Approval to Action? Request updates on this topic by email, or subscribe to Fide’s broader essays and research notes.

Request topic updates →

Opens a draft in your email app. Send it to request updates.

Subscribe to all Fide updates on Substack ↗

Contribute to the research

Bring your perspective.

  • A research lead and enterprise workflow practitioners
  • Operators or auditors who can validate decision boundaries
  • Agent engineers to help design reproducible handoffs and recovery cases
Discuss a contribution ↗

Opens our research inquiry form with this topic included.