Agent requests action
An agent requests an action in a downstream system.
AI execution authority
Who authorized
this exact execution?
AI agents can reach the system. Access does not establish authority for every action they attempt.
Follow the executionExample execution request
PAYMENT.RELEASE.0084
Execution held
- Reason
- Authority not recorded
- Required response
- Named owner approval
- Downstream state
- Payment unchanged
- Next
- Verify or reject
Jochanni Labs
Jochanni Labs is building DAL-X, the Decision Authority Layer. It stops consequential AI agent actions until authority is recorded and verified.
Access controls show what an agent can reach. Logs show what happened. DAL-X checks who authorized the action before it changes a downstream system.
The enforcement boundary
Check authority before the action runs.
DAL-X intercepts
DAL-X holds the action at the enforcement boundary.
Authority is evaluated
DAL-X checks identity, scope, evidence, freshness, and approval.
DAL-X returns a decision
The user sees the decision, reason, required response, and next step.
No authority, no execution.
Primary product
DAL-X
DAL-X checks authority before an AI agent acts. Missing or invalid authority stops the action in any downstream system that verifies DAL-X authority.
- For
- Enterprises using agents to spend money, change records, or affect operations
- Controls
- The point between agent output and downstream execution
- Returns
- Decision state, exact reason, required response, and what happens next
Before DAL-X enforcement
Choose What To Investigate. Review What Is Running.
CAIO-X
Visit CAIO-X- Problem
- The business problem is known, but what to investigate is not.
- Owner
- The executive deciding where AI work should begin.
- Output
- A ranked Investigation Brief.
- DAL-X handoff
- Identifies what to investigate before runtime enforcement.
Decision Governance Review
See the review- Problem
- Agents are approved once and never looked at again. Prompts, models, owners, and policies change while the agent keeps running.
- Why
- Keep agents accountable over time. Check that each still does only what was approved, meets the business need, and is worth running.
- Owner
- The agent owner, analyst, and named executive sponsor. The analyst recommends; the executive sponsor decides and signs.
- Output
- A baseline review with one signed decision per in-scope agent: Keep, Downsize, Replace, or Retire. Checkups follow every 60 days, or sooner when something changes. Each ends in a signed decision and a next date. You receive action rules, a change list, a signed Governance Manifest, and the full decision history. Conducted under mutual NDA.
- DAL-X handoff
- Engineering loads the signed rules into DAL-X, proves that calls without valid authority fail, and updates the rules when decisions change. A signed decision alone is not enforcement. Automatic drift detection is not promised before agents run through DAL-X.
Who owns the decision
Who decides whether the agent can act?
The executive responsible for the workflow sponsors the control. The CIO or Chief AI Officer may own the budget. Security, risk, compliance, legal, and internal audit may require enforcement.
Triggers
- Rogue agent incident
- Board concern
- Regulatory review
- Internal audit finding
- EU AI Act readiness
- Production launch
- Disputed execution
Consequences
- Financial exposure
- Customer harm
- Investigation time
- Production interruption
- Delayed deployment
- Agent suspension
Paid enterprise pilot
One execution path. Thirty days.
Bring one running or deployment-ready agent, its responsible owner, and access to one consequential workflow. The pilot tests authority checks for one action in one downstream system.
01
One agent
02
One consequential action
03
One downstream system
04
Thirty days
You bring
The agent, workflow owner, consequential action, and downstream environment.
We check
Whether the agent has authority when it attempts the action.
The pilot proves
Whether DAL-X can hold the action, explain why, verify authority, and release it.