Policy gates
Policy decides whether a recommendation is allowed, blocked, approval-required, or escalated before action is possible.
- Approval thresholds
- Risk classification
- Policy explanations
Capability
Governance
and human control
Yedron includes control from the start: Decision proposes, Policy gates, Workflow records approval, Action executes, and Audit preserves evidence.
Authority stays visible
A recommendation cannot approve itself.
The control path makes the decision, policy result, human authority, and action request distinct records.
- Proposal
- Reroute shipment to an alternate carrier.
- Policy
- Approval required above the configured cost threshold.
- Authority
- Operations lead reviews the evidence and approves or rejects.
- Evidence
- Decision, policy, approval, and action remain linked.
Illustrative workflow view. Controls and integrations depend on deployment.
Approval record
Workflow records who approved, rejected, overrode, or escalated a decision so accountability stays separate from recommendation logic.
- Reviewer responsibilities
- Override paths
- Escalation records
Audit evidence
Decisions, policy results, approvals, actions, retries, failures, and sensitive reads must leave traceable evidence.
- Decision snapshots
- Execution records
- Access audit
In practice
Make control part of the operating work.
The policy result, accountable owner, and review history should travel with the decision. Define the cases that may proceed and the cases that need human judgment.
Approval thresholds
Route a proposed response for review when its cost, risk, or scope meets the workflow threshold. Show the reviewer the evidence and the policy explanation alongside the requested action.
Exceptions and overrides
Give operators an explicit path to reject, escalate, or override a proposed response within their authority. Keep the reason and review history available to the next owner.
Accountability review
Reconstruct a decision from its source context through the policy result, approval, action attempt, and observed outcome. Keep each responsibility identifiable during review.
How it works
Establish authority before execution.
Governance works best when teams define it with the workflow itself. The control path should remain clear during normal operation and when something goes wrong.
- 01
Define the permitted scope
Agree the decision types, available actions, account boundary, and reviewer roles. Identify thresholds and the evidence needed to evaluate them.
- 02
Apply policy and review
Record whether the proposal is allowed, blocked, or approval-required. Where human authority is needed, preserve the reviewer decision and any escalation.
- 03
Keep the evidence connected
Follow the authorized action, failure path, and outcome. Preserve the records required to explain what happened and why the work was allowed to proceed.
Common questions
Understand the fit.
Must every action have manual approval?
The control requirements are defined for the workflow. Policy determines the gate and required authority; cases that require a human decision are routed through the recorded approval path.
Can an AI recommendation grant authority?
No. A recommendation supplies a proposed response. Policy and Workflow determine whether and how it may proceed, separately from the reasoning that produced it.
What happens when a proposal is blocked?
It does not proceed into execution through that path. The workflow should expose the policy explanation and define whether the next step is correction, escalation, or closure.
Start where one workflow needs clearer control.
Adoption starts with one workflow, the systems around it, the approvals it needs, and the outcome your team wants to improve.
Design a decision loop