HomeThe ProblemHow Protection WorksProofPilotAboutDiscuss a controlled pilot
Controlled Pilot

Start with one consequential workflow.

A Josephine pilot is deliberately bounded. One company-selected workflow. One approved protective boundary. One supervised test of whether Josephine can travel with the work, detect deterioration and protect only within the authority your company provides.

A deliberately simple 1-2-3 front door.
The 1-2-3 Pilot

A simple front door to a controlled process.

1

Choose the workflow.

Select one consequential AI-driven workflow where deterioration could create a meaningful business consequence. You define what the work is and why it matters.

2

Connect it.

Connect Josephine beside the workflow to the company-controlled records that establish what is happening, what protection may be requested and what happened afterward. Where the required evidence and connector interfaces are compatible, Josephine is designed to begin beside the workflow without requiring redesign of the underlying AI system or an explanation of every agent before beginning..

3

Approve the boundary and run the supervised test.

Your company reviews the scope and defines what Josephine may and may not do. The supervised test then verifies identity, evidence, authority, protection and observed effect before pilot operation proceeds.

Your company defines the boundary. Josephine stays inside it.

The purpose of the pilot is not to hand Josephine open-ended control. It is to establish a narrow, evidence-backed protective role around one consequential workflow.

What the Company Provides

Josephine does not invent the facts it needs.

Business owner

Define the work and consequence.

  • The workflow Josephine is being asked to protect.
  • What deterioration means in business terms.
  • What consequence matters and when management must own the next decision.
  • What protection the company is willing to authorize.
Technical representative

Point Josephine to the evidence.

  • Where workflow events can be observed.
  • Where an authorized protection request may be sent.
  • Where requested, acknowledged, applied and consumed evidence can be verified.
  • Where the effect of protection can be observed afterward.
Connecting an evidence source does not prove an event occurred. Josephine keeps the evidence stages separate and does not turn a request, acknowledgement or application into proof of the next stage.
What Happens During the Pilot

Observe first. Verify before trust. Protect only inside authority.

1 · Establish
Workflow identity, work-item identity, evidence sources and company authority are established. Missing material facts are not silently filled in.
2 · Observe
Josephine begins without assuming intervention is needed. The supervised setup verifies that the correct work is being seen and that company operations remain independent.
3 · Verify
A harmless control or sandbox equivalent is used where appropriate. Requested → Acknowledged → Applied → Consumed → Effect observed are verified as separate evidence stages.
4 · Protect
If supported deterioration occurs, Josephine may initiate only the protection the company authorized. It does not create new permission because a situation looks serious.
5 · Management
Protection is intended to create time for people to understand and react. Developing conditions may remain in an Observe / amber state; conditions requiring management attention move to the designated escalation path.
6 · Recover
Josephine continues watching after intervention. Recovery, recurrence, uncertainty and the end of delegated authority remain visible rather than being hidden by a simple “protected” label.
Company Control

The pilot does not transfer the consequential decision to Josephine.

Josephine can operate only within the authority your company has explicitly provided for the selected workflow. When that authority ends, the next consequential decision belongs to management.

Protect the work. Buy management time. Preserve management ownership of the consequence.

Evidence Boundary

The first company pilot is supposed to prove something new.

Josephine's engineering/package foundation has been verified for a supervised pilot with one authorized workflow. What has not yet been established is real-company compatibility, real company-supplied evidence and authority, or real-world operational effect.

That distinction is intentional. A controlled pilot is the next source of real-world evidence, not a ceremony around claims we already assume to be true.
Discuss a Controlled Pilot

Could Josephine protect one of your consequential AI-driven workflows?

Tell us enough to understand the workflow and why it matters. The next step is a conversation about suitability, evidence and the protective boundary.

Josephine is being introduced through controlled supervised pilots, not public self-service deployment.

Pilot inquiry

Keep it brief. We only need enough information to decide whether a controlled discussion makes sense.

Protected against automated abuse without a visible CAPTCHA or third-party challenge.
No account will be created. By submitting, you acknowledge the Privacy Policy and Terms.