HUMANLATCH by VerdictLayer

Interactive simulator

Interactive workflow demo

See the full approval loop, not just the final verdict.

Pick a scenario, play the workflow, and compare what happens with and without HumanLatch between the agent and execution boundary.

Scenario presets

Compare

HumanLatch intercepts the action, scores it, matches policy, and decides whether to approve, escalate, or block.

Step 1

Agent proposes a consequential action

Step 2

Risk engine scores the request at 50

Step 3

Policy matched: Require approval for production changes

Step 4

Human review is requested

Step 5

Execution waits for approval

Workflow map

Agent proposes a consequential action

The action is intercepted before it can touch a live system.

Agent

Proposes action

HumanLatch

Scores + matches policy

Reviewer

Approves or rejects

Execution

Runs or blocks

Current action

Apply Terraform changes to production EKS cluster

terraform.applyprod-eks-cluster

Pending approvalRisk 50

Live event timeline

09:41:12

proposal.received

Agent proposes a consequential action

09:41:13

risk.scored

Risk engine scores the request at 50

09:41:14

policy.matched

Policy matched: Require approval for production changes

09:41:16

approval.requested

Human review is requested

09:41:21

execution.pending_approval

Execution waits for approval

Customize action

Verdict summary

Pending approvalRisk 50

This action needs a human sign-off before execution should continue.

Risk factors

Production context

+30

Production actions get elevated review pressure.

Sensitive target

+20

The target suggests production or admin scope.

Policy match

Require approval for production changes

Outcome: require_approval

Priority: 100