When AI Acts, a one-day working session for business leaders. Wednesday, January 13, 2027, Irvine, California. Early bird $395 until December 18.

See the day and register

Engagements that put the model to work

Each engagement delivers one part of the Three-System Model. Most clients start with the executive workshop and build from there.

Executive workshop

Half a day, your leadership team, from $7,500. All three systems, at decision level.

A working session at your offices for the executive team of an organization piloting or running AI agents. The team maps its agents, tests its decision rights in a live tabletop exercise, and designs governance for its most important workflow.

You receive

  • An exposure map of your agents, with the three workflows to govern first
  • Decision rights tested under pressure
  • A one-page governance design and a 90-day plan with owners

The workshop in detail

Agentic Safety Audit

Every agent, mapped and scored. Exposure across all three systems.

An inventory and assessment of every agent and AI workflow in scope, including the ones that arrived inside other software. Each is scored against eight executive questions covering purpose, authority, exposure, prevention, intervention, inquiry, provenance and defensibility.

You receive

  • An agent inventory
  • An exposure map by impact and autonomy
  • A gap report organized by system, with priorities

Trust Harness

The controls, designed and tested. System 1 and System 3: intent, authority, prevention and intervention.

The design and testing of the controls that keep agents inside the authority you gave them. Limits are enforced outside the model, someone can always pause an agent, and oversight is measured rather than assumed.

You receive

  • Intent specifications, versioned and checked by an intent pre-mortem
  • A delegation model and pause rights for each agent
  • Continuous Inquiry, with the detection rate of each mechanism tested

Evidence Fabric

The records that prove it. System 3: substantiate.

Making control provable. Decision records are anchored to what agents did, kept to a standard an examiner, auditor or court can follow, and available before anyone asks for them.

You receive

  • A decision provenance design
  • Record standards for agents and the people who oversee them
  • Reporting ready for examiners, auditors and the board

How we work

Vendor-neutral

We work with whatever agents and platforms you run.

Claims you can test

We do not offer guarantees. We tell you what each control should catch and how to measure whether it does.

Alongside your counsel

We work with your general counsel, not instead of them. Nothing we provide is legal advice.

Evidence over narrative

Records made at the time carry more weight than accounts written afterwards, so we build for the record.