davisAIgovDavis & Associates Advisory, LLC
Connect with us

AI Harness Explorer

Follow a proposed action from model output to the system that can execute it. Ask which identity is acting, what policy permits and which record shows the downstream result.

Open the interactive architecture

Use the explorer in two minutes

  1. Open the overview and choose a trust zone.
  2. Follow one inference from input to a downstream action. Inspect decision authority and data access at each boundary.
  3. Identify the record that would allow a reviewer to reconstruct the action later.

For the complete component catalogue and its video walkthroughs, see the AI Harness capability matrix.

Architecture walkthrough

Control boundaries

Where AI actions are authorized and recordedInputs reach the model, which proposes an action. A policy check evaluates approved scope. The tool or its controlling service enforces the decision. The policy decision and downstream action record feed the evidence store.InputProposalPolicy decisionEnforcementEvidence

Examine the telemetry evidence

Compare the operating models

Ungoverned

A model can call a tool directly, and the record shows only the resulting action.

Governed

A scoped identity and policy gate authorize the tool call. The record joins the decision, approved scope and actual outcome.

Policy configuration and an illustrative authorization check

YAML and JSON show release settings. Python illustrates the reviewer-role and approval checks. The application must obtain those inputs from trusted services, bind approval to the action and record, enforce the decision at execution, and record both the decision and outcome.

YAML

action: release_record
allowed_role: quality_reviewer
require_approval: true
record_decision: true

JSON

{
  "action": "release_record",
  "allowed_role": "quality_reviewer",
  "require_approval": true,
  "record_decision": true
}

Python

def permits_release(role, approved):
    return role == "quality_reviewer" and approved

Discuss an AI Harness assessment