Jochanni Labs · Decision Governance
DAL-X holds a consequential AI agent action until a person with authority decides, and writes a tamper-evident authority record. It has run end to end on real systems in a sandbox. There is no customer deployment yet.
Enforcement Loop (Illustration)
Agent Output
intercepted at intake
Execution Blocked
default state
Review Queue
human reviewer
Authority Granted
decision recorded
Gate Cleared
token validated
Execution
receipt written
The Problem
Illustrative examples of what an agent could be trusted to do. Without an authority check, the agent logs its decision and the action executes. Nobody authorized it first. A log is not an authority record.
The agent logged the decision. That is not the same as a human authorizing it.
The Solution
DAL-X sits between AI agents and downstream execution systems that enforce DAL-X token verification. Execution is blocked by default. On those paths, authority is required before a consequential execution proceeds.
DAL-X
Enforcement Gate
Execution blocked by default
Auto release is suspended. Every submission waits for a human authority decision before it can be released.
Four reviewer actions
Approve, Reject, Escalate, or Request revision. Each requires a reason and writes a tamper-evident decision record before execution status can change.
Token required at the gate
The downstream system verifies DAL-X authority before it executes, by calling the DAL-X gate or by verifying the signed token itself. No valid authority. No execution. The gate applies to execution paths that enforce DAL-X token verification.
Jochanni Labs · Decision Governance
DAL-X has run end to end on real systems in a sandbox: a real agent and a separate downstream system that refuses any action without DAL-X authority. There is no customer deployment yet. The next step is a first enterprise case against a real workflow.