Authority Beyond the Role

Overview

Everything so far in this module happens in rooms you are in. This topic is about influence that operates while you are elsewhere.

The claim: an architect who publishes reasoning accumulates influence faster than one who only makes decisions — and the mechanism is not visibility or self-promotion. It is that written reasoning gets reused by people you never spoke to.

The mechanism

A decision reaches the people in the conversation. A recorded rationale reaches everyone who later faces the same question.

Which produces a compounding effect that a decision does not have:

• A new engineer reads why the boundary exists instead of removing it as an inconvenience.
• A team facing a similar choice applies your criteria without asking you.
• Someone proposing a rejected option finds the reasoning already there, and either accepts it or brings a genuinely new argument.
• A colleague in another department cites your framing in a decision you were not party to.

Section 1’s size topic argued that at a few hundred engineers the single architect becomes the bottleneck. Published reasoning is the mechanism that defers that — it lets your judgement operate without your presence, which is the only form of scale available to one person.

Four channels, in order of leverage

graph TD
  A["<b>Decision records</b><br/>why, not what<br/><i>highest leverage,<br/>lowest effort</i>"] --> E["Influence that<br/>operates while<br/>you are absent"]
  B["<b>Internal writing</b><br/>criteria, reviews,<br/>post-decision notes"] --> E
  C["<b>Teaching and<br/>speaking internally</b>"] --> E
  D["<b>External writing<br/>and side projects</b><br/><i>slowest, real</i>"] --> E

  classDef best fill:#7D9A4E,stroke:#5F7A38,stroke-width:2px,color:#ffffff
  classDef mid fill:#5B8FB9,stroke:#4A7A9E,stroke-width:2px,color:#ffffff
  classDef slow fill:#E8A33D,stroke:#C4832A,stroke-width:2px,color:#ffffff
  class A best
  class B,C mid
  class D slow
  classDef out fill:#D9695B,stroke:#B4483A,stroke-width:2px,color:#ffffff
  class E out

Four channels. The olive one is first because it costs almost nothing and is read by exactly the people whose decisions you want to influence.

Decision records. The highest-leverage channel and the cheapest. The content that matters is why, not what — the constraint that forced the choice, the options rejected, the condition that would reopen it. A record of what was decided has modest value; a record of why is what stops the decision being undone by someone who cannot see the reason.

Internal writing. Criteria catalogues, review notes, short post-decision write-ups. The useful property is that these are read by people deciding things you will never hear about, which is precisely the influence you cannot otherwise have.

Teaching and speaking internally. A thirty-minute session on how to evaluate a class of decision propagates a method rather than an answer. Slower than writing to produce, and it builds relationships that writing does not.

External writing and side projects. Genuinely builds authority, and slowly. Its main internal effect is indirect: being visibly competent outside the company changes how your judgement is received inside it. Treat it as a long investment rather than a tactic, and be honest that its payoff is uncertain.

What this is not

It is not self-promotion. Writing that exists to make you look good is detectable and counterproductive. The test is whether the artifact is useful to someone who does not care about you — a criteria catalogue is; a summary of your achievements is not.

It is not volume. This module’s trust topic established that talking more does not persuade more, and the same applies in writing. One clear decision record beats a weekly newsletter.

It is not a substitute for the conversations. The pre-socialising from Section 1 still has to happen. Published reasoning compounds influence you already have; it does not create it from nothing, because an unknown author’s document is read as an opinion.

The honest costs

It takes time you do not obviously have. Writing a decision record properly is an hour, and the week always contains something more urgent. The only reliable defence is making it part of the decision rather than a follow-up — the decision is not finished until the record exists.

It makes you auditable. Written reasoning can be checked against outcomes, and some of yours will be wrong. That is the point: Section 1 established that visibly withdrawing a wrong claim builds credibility. But it is uncomfortable, and an architect who writes nothing is never caught out.

The payoff is delayed and diffuse. You will rarely see the decision your document shaped. The influence is real and mostly invisible, which makes it hard to sustain without believing the mechanism.

What to publish first, and why

With no mandate, no practice of recording decisions, and about 140 engineers who have never had an architect, the question is what single thing to publish.

Not a standards document. Nobody asked for it, nothing enforces it, and it would read as an attempt to acquire authority — the antipattern from Section 2’s remit topic.

Decision records, starting with the ones already made. Four exist from Modules 4–7 and each carries reasoning that will otherwise be lost:

1. Hosted in-region inference. Why, what was rejected, and the reversal condition — in-region processing not contractually guaranteed, or cost above $2.00 per interaction.
2. Two services rather than one. The retrieval and action paths have different NFR budgets and their expected changes do not cross. This is the record that matters most, because Section 1’s Conway analysis predicts the boundary will erode within a year if one team owns both — and a future engineer who finds a reason may preserve it, where one who finds an inconvenience will not.
3. Retrieval rather than fine-tuning, with the reconsideration trigger for a narrow high-volume classifier.
4. The regression gate and its arithmetic, including why a three-point trigger on a single run could not have fired. Attributed to Ivan Petrov’s review, which is the public credit from the sceptic topic doing double duty.

Why these four are the right first publication. Each answers a question somebody will ask. None claims authority. Each is checkable. And collectively they establish a format — context, decision, consequences, reversal condition — which is how a practice starts without anyone having to approve a practice.

Module 9 turns that into a governance proposal. The point here is that it begins as four documents nobody authorised, written because the decisions were already made and the reasoning was going to evaporate.

Feeds the workbook

Your justification pack is the first instance of this channel. Write it so that someone who was not in the room — and who will arrive after you have gone — can reconstruct the decision and tell what would reopen it.

Check your understanding

1. Name a decision in your own work whose reasoning exists only in your head. What happens when you leave?
2. Which of the four channels is cheapest for you this month? What would you write?
3. Your published reasoning turns out to have been wrong. What do you do, and what does Section 1 say that buys you?

---

→ That closes the module. You can read the organisation and make a decision legible to it. — Module 9 — Making the Practice Repeatable

Optional workbook: Exercise 8 turns this module into artifacts. The exercises build on each other, so work them in order or not at all.