Why is self-orientation a denominator and not a subtraction?

Because its effect is proportional. It scales down everything in the numerator at once, which is why high-competence leaders with visible self-interest often score worse on trust than modest ones without it.

How do I measure intimacy without making it awkward?

Do not ask about it. Look at behavior. Count how often someone brings you a problem before it is unavoidable. That number is the measurement.

What actually stops a deepfake attempt?

A pre-agreed verification protocol and a culture where questioning an urgent instruction from a senior person is safe. The Ferrari case turned on both at once.

Is the Trust Equation applicable to AI outputs?

The framework was built for human trust and the four components extend usefully to machine outputs: is it credible, is it reliably so, can you interrogate it, and whose interest is the system optimized for.

Where do most leaders actually lose trust?

Reliability, through quiet misses, and self-orientation, which they cannot see. Credibility is usually the component people worry about and rarely the one failing.

Part III · Strategic Leadership in the AI Era

The Trust Equation

The Trust Equation puts credibility, reliability and intimacy in the numerator and self-orientation in the denominator.

What it diagnoses

Trust treated as a feeling instead of a structure with four measurable inputs.

The Trust Equation puts credibility, reliability and intimacy in the numerator and self-orientation in the denominator. That last position is the whole framework. Self-orientation does not subtract from trust, it divides it, which means a leader can be credible, dependable and genuinely safe to talk to and still not be trusted, because everything they do is visibly in service of themselves.

Run as a diagnostic, each component fails distinctly.

Credibility is whether people believe your claim and believe you can deliver it. It is about what you say.

Reliability is whether you do it consistently. It is about what you do, and in distributed teams it is the component most often quietly failing, because inconsistency is invisible without proximity.

Intimacy is not closeness. It is psychological safety: whether people will bring you a concern, admit a mistake, or ask something that might sound uninformed.

Self-orientation is whether your attention is genuinely on the other party’s interests. It is the only component that divides, and it is the one almost nobody audits about themselves.

The AI era adds a specific pressure at the top of the list. Credibility has historically been established by presence, voice and face. Those can now be synthesized, which means credibility needs a verification mechanism that does not rely on recognition.

How to run it

The book runs this as a four-component system.

Build credibility deliberately. Make claims that can be checked, and make the checkable ones often enough that people stop needing to check. In the deepfake era, add an explicit verification path: an agreed channel or challenge for any high-consequence instruction, especially anything involving money or access.
Systematize reliability. Reliability is a systems property, not a character trait. Commitments logged, deadlines tracked, and a standing practice of renegotiating early rather than missing quietly. One early renegotiation costs less trust than one silent miss.
Develop intimacy structurally. Psychological safety is produced by what happens to the first person who brings bad news. The practice is to make surfacing a problem visibly safer than concealing it, which is a decision about consequences rather than a matter of tone.
Reduce self-orientation. The audit question: in the last ten interactions, whose interest was served by how I framed things. This is the hardest component to see from the inside, which is why the framework pairs it with external feedback rather than self-assessment.

How operators misuse it

They work only on the numerator. Credibility and reliability are the components with obvious practices attached, so they get the attention. A leader who improves both while self-orientation stays high has multiplied a bigger number by the same fraction and moved nothing.

They read intimacy as rapport. Being liked and being safe to tell the truth to are different properties and they come apart routinely. A warm leader whom nobody brings bad news to has rapport and no intimacy.

They treat verification as an IT problem. Deepfake-enabled fraud targets people, through urgency and authority, not systems. The control is a human protocol agreed in advance, and it belongs to finance and leadership rather than security alone.

What the pattern shows

In July 2024, a Ferrari executive received messages and then a call from someone using a convincingly synthesized version of CEO Benedetto Vigna’s voice, accent included, pushing an urgent and confidential deal. The attempt failed at a single point. The executive asked which book Vigna had recommended to him a few days earlier, and the caller could not answer.

Every technical component of credibility was successfully forged: voice, manner, urgency, apparent authority. What held was intimacy in the framework’s precise sense. There was enough real relationship to hold a shared fact that could not be scraped, and enough safety in that relationship for a subordinate to interrupt an urgent instruction from the CEO with a test.

The research describes the same pattern at scale. Deloitte’s work on generative AI trust standards documents how quickly synthetic media is eroding the reliability of what people see and hear, and cites analyst estimates that the deepfake detection market will grow 42 percent a year, from $5.5 billion in 2023 to $15.7 billion in 2026. Detection spend is rising because recognition has stopped being evidence.

A 2025 Harvard Business Review article on AI trust adds the organizational half, and it is the more useful finding for an operator. Employee skepticism about AI tracks their confidence in leadership. Trust in the tool is not separable from trust in the people deploying it, which means an AI rollout inside a low-trust organization is not a technology problem waiting for better change management. It is the denominator.

Read together: credibility is now forgeable, reliability and intimacy are not, and organizations that invest only in detection are defending the one component that has already fallen.

Where it lives in the book

Framework 19, Chapter 6, page 369. The Integrity System. It closes Chapter 6 and closes the book, which is deliberate: every framework before it assumes the organization will act on what it produces, and the Trust Equation is what determines whether it does.

Questions operators ask

Why is self-orientation a denominator and not a subtraction?: Because its effect is proportional. It scales down everything in the numerator at once, which is why high-competence leaders with visible self-interest often score worse on trust than modest ones without it.
How do I measure intimacy without making it awkward?: Do not ask about it. Look at behavior. Count how often someone brings you a problem before it is unavoidable. That number is the measurement.
What actually stops a deepfake attempt?: A pre-agreed verification protocol and a culture where questioning an urgent instruction from a senior person is safe. The Ferrari case turned on both at once.
Is the Trust Equation applicable to AI outputs?: The framework was built for human trust and the four components extend usefully to machine outputs: is it credible, is it reliably so, can you interrogate it, and whose interest is the system optimized for.
Where do most leaders actually lose trust?: Reliability, through quiet misses, and self-orientation, which they cannot see. Credibility is usually the component people worry about and rarely the one failing.

Sources

Score it

Take the Burnout Diagnostic Get The Resilience Engine: $27

Pairs with

Human-in-the-Loop. Human-in-the-Loop filters the AI output, the Trust Equation filters which output the organization will act on.
Async-First Default. The Trust Equation supplies the governance architecture, Async-First supplies the operating protocol.
Clear is Kind. Reliability and intimacy both depend on the standard having been stated plainly in the first place.

Written about

Human Oversight of AI: The Signature Is Not the Check