Earning Trust from the Sceptic

Overview

Section 1 located the person who can stop your initiative without any authority to do so, and established that their scepticism traces to a specific prior failure.

This topic is the conversation. It is the highest-leverage hour available to you, and almost everything intuitive about it is wrong.

Their scepticism is information, not resistance

The reframe that makes the rest work.

Section 1 established that in an organisation where a failure could not be discussed openly, scepticism is the only available form of risk reporting. The person objecting is not being conservative or territorial by default — they are carrying information about a failure mode the organisation declined to examine, in the only form available to them.

Which means the objection has content, and extracting it is more valuable than overcoming it.

It also means the usual framing — how do I get them on side? — sets the wrong goal. You are not trying to convince them the system will work. You are trying to make its failure mode detectable, which is a claim they can verify rather than a promise they must trust.

Five moves

graph TD
  A["<b>1 · Go first, privately</b><br/>before any group"] --> B["<b>2 · Ask what they<br/>would need to see</b><br/><i>not 'do you agree'</i>"]
  B --> C["<b>3 · Give the mechanism,<br/>not the reassurance</b>"]
  C --> D["<b>4 · Let them set<br/>the number</b>"]
  D --> E["<b>5 · Credit them<br/>publicly</b>"]

  classDef early fill:#5B8FB9,stroke:#4A7A9E,stroke-width:2px,color:#ffffff
  classDef core fill:#7D9A4E,stroke:#5F7A38,stroke-width:2px,color:#ffffff
  classDef last fill:#E8A33D,stroke:#C4832A,stroke-width:2px,color:#ffffff
  class A,B early
  class C,D core
  class E last

Five moves in order. The second is the one that changes the conversation; the fourth is the one that makes the agreement durable.

1 — Go first, privately. Section 2’s sequencing rule, and its most important application. An objection raised in a group becomes everyone’s doubt, publicly unanswered, with the room having watched you be surprised. The same objection a week earlier, privately, becomes a design change you present as your own thinking.

2 — Ask what they would need to see. The move that changes the conversation. Not do you agree? — which invites a verdict — but what would make you comfortable that this will not go the way the last one did?

This converts opposition into specification. You get a list of conditions, which is a vastly better artifact than a yes. And it is a question they can answer honestly without conceding anything, which makes it easy to say.

3 — Give the mechanism, not the reassurance. “It will be different this time” is what everyone says. What works is the structural answer: the last one degraded without anyone noticing; here is what measures it, here is the threshold, here is what happens automatically when it is crossed. Section 2’s scar-tissue topic argued this is the only credible response, and the reason is that it is checkable.

4 — Let them set the number. Module 2’s elicitation technique, cashed in. Their objection becomes the threshold, and people rarely veto a number they chose. This also transfers something real: they now have a stake in the safeguard working rather than in the initiative failing.

5 — Credit them publicly. The safeguard is theirs, and saying so in the room costs nothing. It converts a sceptic into a co-author, which is a far more stable position than a sceptic who was persuaded.

What not to do

Do not argue the technical merits harder. The objection is not primarily technical, so winning on technical grounds does not resolve it — and winning an argument against someone with informal power is usually a net loss.

Do not go around them. Escalating to their manager converts a recoverable disagreement into a permanent one, and Section 1 established that their power does not run through the reporting line anyway.

Do not characterise them as resistant. Naming someone as an obstacle, even privately, is unprovable and corrosive.

Do not promise it will be fine. You cannot know that, the previous topics established what unfounded confidence costs, and this person in particular has heard it before.

The honest limit

Not every sceptic is carrying information. Some are protecting territory, or are simply opposed to change.

A workable diagnostic: does their objection name a mechanism or a conclusion?

“Nothing will measure whether this is still working” names a mechanism. It is specific, checkable, and addressable — and it is what a well-founded objection looks like.

“This will not work here” names a conclusion with no attached reasoning. Ask twice for the mechanism; if none arrives, the objection is positional rather than informational, and the response is Section 2’s: a mechanism in the open — explicit criteria agreed before the decision — rather than a conversation.

Most of the time, in my experience of how these situations are described, the objection turns out to have content. Treat that as the default and the positional case as the exception, because the cost of getting it wrong in that direction is much lower.

The conversation with the monolith’s owner

Ivan Petrov: principal engineer, eight years, owns the Rails monolith that still runs most of the business, widely respected, no formal veto, and able to stop this.

What he knows. In 2023 the company shipped a machine-learning feature for smart routing. It worked at launch, degraded over months, and was switched off quietly with no announcement and no postmortem. His rational inference: this organisation cannot detect slow quality decay and does not examine it afterwards.

The conversation, in order.

Privately, before the leadership presentation. Not after it, and not in it.

The question: what would you need to see to believe this will not go the way smart routing did? Expect a specific answer, because he has thought about it for two years.

The mechanism, prepared in advance: 300 curated questions measured weekly, a correctness floor, a confident-error ceiling ten times tighter than the floor, and a release gate that blocks automatically rather than requiring a person to raise a concern. Section 1’s culture reading is why that last property matters — in an organisation where the failure could not be discussed, a mechanism needing no messenger is the only reliable form of risk reporting.

The detail that will actually move him: the regression trigger as originally specified was a three-point drop measured on a single run, and the arithmetic says that sits inside the measurement’s own noise — roughly ±5.8 points at 300 cases run once. So it was corrected to four repetitions, giving about ±2.9. That is evidence the safeguard was checked rather than asserted, and it is the single most credible thing you can bring him. Anyone can propose a gate; almost nobody verifies that theirs could fire.

His number: he sets the regression threshold. Module 2 already named him as reviewer on the confident-error ceiling for this reason.

Public credit: in the leadership pitch, the regression gate is attributed to his review. It costs nothing and it changes what he is defending.

Why this is the highest-value hour available. He is not in the minimum winning coalition — Maya, Dev, Lena and Tomas are. He can stop it anyway. And the thing that moves him is not persuasion; it is arithmetic about whether the safeguard works.

Feeds the workbook

The rubric requires the prior-failure objection answered before it is asked, which is a pitch requirement built on this conversation. Your justification pack should record which threshold the sceptic set, because an artifact showing that is more persuasive to everyone else than any argument you could write.

→ That closes trust. The decision itself still has to be defensible on paper. — The comparison table