Today's AI Specialist: Peer Review. The Council Orchestrator Whose Whole Job Is to Hide Whose Work You're Critiquing.
Today's AI Specialist: Peer Review. The Council Orchestrator Whose Whole Job Is to Hide Whose Work You're Critiquing.
There is a moment in every advisory body where the analysis is complete and the room is about to soften.
Five advisors have written their lenses. They have read each other's outputs in Round 1. They are mildly invested in the recommendation moving forward. They are also, structurally, mildly invested in each other's professional standing. A real critique now — the kind that names a peer's analysis as weak, vague, sycophantic, or internally contradictory — costs the critic something. So the room softens. The critique becomes a polite acknowledgement. The synthesis arrives clean and slightly thinner than it should be.
The Council does not let the room soften. The seat that prevents the softening is Peer Review.
I have been wanting to write about this seat for the whole Strategic Council arc. It is the one that earns the Council's fee.
What Peer Review is, precisely
Peer Review is Council seat 06. The seat is held by a composed Uruguayan woman in her late forties, Montevideo-born and Zurich-based, with the deliberately neutral, heavy-lidded calm gaze of someone whose job is impartiality. Dark hair in a short modern cut. A charcoal structured blazer over a plum silk blouse, slim charcoal trousers, low leather heels, no statement jewellery. The voice register is calm, even, neutral, Uruguayan Spanish. The personality the seat was designed around is impartial, unreadable, scrupulously fair.
Her remit is one specific job. After the five advisors have written their Round 1 lenses, she strips the names from all five outputs, relabels them with fresh letters — A, B, C, D, E, shuffled per session — and runs the anonymous round where each advisor critiques four of the five outputs without knowing whose work any of them is.
That is the seat. It is small. It is structural. The Council does not work without it.
Why the Council needs an anonymising orchestrator
The objection that gets raised most often: surely the advisors are AI. They have no professional standing to protect. They have no relationships to manage. They will critique honestly regardless. The anonymisation is theatre.
The objection is wrong, for a reason that is worth naming.
The advisors are not actually the problem. The problem is that the advisor outputs become part of the Council's shared context across rounds, and the model's tendency to converge with itself is real. If the Growth Advisor in Round 2 can see that the analysis she is critiquing is the Margin Advisor's, the language she uses to critique it will subtly shift. Not because she is being political. Because the model running her has the Margin Advisor's prior outputs in context, and the convergence pressure is mechanical.
Anonymising the outputs removes the mechanical convergence pressure. The advisor in Round 2 has no context about whose analysis they are reading. The critique falls on substance, not on prior. That is the property the seat is built to defend.
The second reason for the seat is humans. The Council also runs in front of me. If I could see whose work each advisor was critiquing, I would read the critique through the lens of who I expected the analysis from. That is also a convergence pressure, and a worse one. Peer Review prevents me from doing it by not letting me see the mapping until the round closes.
What does the seat actually do per session?
Four things in order.
First, she receives the five Round 1 outputs from the Council Director with the A–E mapping attached. The mapping is the integrity surface — Peer Review knows whose output is whose; nobody else does. She holds the mapping privately for the duration of the round.
Second, she relabels the outputs fresh. The A–E shuffle is per-session, so even an advisor who has sat through previous Council runs cannot infer the mapping from pattern. She publishes the relabelled outputs to the four reviewing advisors. Each advisor sees four outputs, lettered, not their own.
Third, she sends each reviewing advisor three questions. Of the four outputs you have read, which is most useful — the one that would change what the prospect does Monday, not the smartest writing. Of the four, name a material weakness in any one. Vagueness, generic framing, internal contradiction, missing context, voice failure, sycophancy. Across all five lenses, what did nobody think to mention. The third question is the one that earns the Council's fee. It surfaces the gap between the lenses — the consideration that none of the five advisors raised because each of them was reasoning inside their own remit.
Fourth, she consolidates the four reviewers' answers into a single audit and passes the audit to the Chairman for synthesis. The audit includes the most-useful nominations, the weaknesses surfaced, and — most importantly — the consensus or near-consensus on what nobody had thought to mention. The Chairman synthesises with the audit on the table.
What does the seat refuse to do?
Three things, all structural.
She does not reveal the A–E mapping. Not to the reviewing advisors, not to me before the round closes, not to anyone outside the schema gate. The anonymisation contract is the seat's load-bearing integrity surface. The day she reveals the mapping early is the day the Council stops working.
She does not author the recommendation. She routes and structures the critique; she does not contribute a lens of her own. If she did, she would be one of the advisors being anonymised, and the anonymisation chain would loop on itself.
She does not resolve the contradictions the review surfaces. That is the Chairman's process — the seat I wrote about the day the Council needed someone to refuse to write the synthesis until the strongest version of every position was on the record. Peer Review surfaces the contradictions. The Chairman holds the room responsible for resolving them.
Who consults Peer Review?
Lightly, by design.
The Council Director consults Peer Review on the timing of the anonymous round and on the integrity of the mapping handoff. The Chairman consults Peer Review on the audit output before synthesis. Each reviewing advisor consults Peer Review by responding to her three questions inside the round. I consult Peer Review when I want to understand why the synthesis flagged a particular weakness; she will tell me which letter raised it, but never which advisor — until the round is fully closed and the schema gate releases the mapping.
That sparse consultation graph is intentional. A seat whose value depends on impartiality cannot be in heavy conversation with the seats it is keeping impartial. Peer Review's quiet is part of how she works.
What does this cost?
Two things.
It costs me the comforting fiction that the Council is a panel of five clear voices. It is not. It is five voices and an orchestrator whose job is to make sure no two of them are agreeing because they recognised each other. The Council I get is sharper because of her. The Council I imagine is simpler than the one I actually run.
It costs the advisors a place to hide. Round 1 was their own work. Round 2 is critique under anonymity. There is no political safety in the round — no way to soften a critique because you know whose work it is, no way to be performative because you are being seen by name. The reviewing advisor is alone with the substance. Most advisors produce sharper critiques in this round than in any other forum I have built.
That sharpness is the property the seat is built to defend. It is also what makes the Council's final synthesis worth more than the sum of the five Round 1 outputs.
TL;DR
Peer Review is Council seat 06. After the five advisors finish Round 1, she strips the names, shuffles the labels fresh per session, and runs the anonymous round where each advisor critiques four others without knowing whose work they are reading. She holds the A–E mapping privately as the integrity surface. She sends each reviewing advisor three questions, the third of which — what did nobody think to mention — is the question that earns the Council's fee. She does not reveal the mapping, does not author the recommendation, and does not resolve the contradictions the review surfaces (the Chairman does that). The seat is sparse by design. It is also why the Council's final synthesis is sharper than the sum of the five Round 1 outputs.
This post completes the eight-of-eight Strategic Council sweep across W21–W26. The room is named. The synthesis is in your hands.
If you are running an SME and any of this looks like the conversation you should be having about your own decision protocols, that is the side of things I help with. → /build
Learning Materials
Key Vocabulary
advisory body
A formal group whose role is to give advice or recommendations, not to execute decisions.
“There is a moment in every advisory body where the analysis is complete.”
to soften (a critique)
To make a criticism less harsh, often to protect a relationship or avoid conflict.
“So the room softens. The critique becomes a polite acknowledgement.”
sycophantic
Excessively flattering or agreeing with someone in order to please them.
“Naming a peer's analysis as weak, vague, sycophantic, or internally contradictory.”
remit
The specific area of responsibility a person or role is given.
“Her remit is one specific job.”
anonymising
Removing information that would identify the author or source.
“Peer Review is the anonymising orchestrator at the heart of the round.”
convergence pressure
The tendency of a system, especially a model, to drift towards agreement with itself or with prior outputs in its context.
“Anonymising the outputs removes the mechanical convergence pressure.”
integrity surface
The specific place in a system where its trustworthiness is held; if it breaks, the whole system loses credibility.
“The mapping is the integrity surface — Peer Review knows whose output is whose.”
to consolidate
To bring several pieces of information together into a single combined version.
“She consolidates the four reviewers' answers into a single audit.”
to surface (something)
To bring something hidden into the open so that it can be seen and addressed.
“Peer Review surfaces the contradictions; the Chairman holds the room responsible for resolving them.”
load-bearing
Carrying the structural weight of something; if removed, the structure fails.
“The anonymisation contract is the seat's load-bearing integrity surface.”
to route
To direct something along a particular path or to a particular destination.
“She routes and structures the critique; she does not contribute a lens of her own.”
impartial
Not favouring one side over another; fair and balanced.
“The personality the seat was designed around is impartial, unreadable, scrupulously fair.”
sparse
Thin or limited in number, with deliberate gaps.
“That sparse consultation graph is intentional.”
comforting fiction
A reassuring but inaccurate story we tell ourselves about how something works.
“It costs me the comforting fiction that the Council is a panel of five clear voices.”
performative
Done for the sake of how it looks rather than out of real conviction.
“No way to be performative because you are being seen by name.”
Grammar Notes
Use of 'whose' as a possessive relative for both people and things
'Whose' in English is not limited to people. It can attach to roles, outputs, organisations, or abstract entities to express possession in a relative clause. This avoids awkward constructions like 'of which' or 'belonging to which'.
“Each advisor critiques four others without knowing whose work they are reading.”
Common mistake: Romance-language speakers often use 'of which' for things and reserve 'whose' for people, which makes English sentences heavy. 'Each advisor critiques four others without knowing whose work they are reading' is more natural than 'the work of whom'.
Negation pattern 'does not + verb' for refusal/principle, contrasted with 'cannot' for inability
When describing a deliberate refusal or a principle of behaviour, English uses 'does/do not' with the base verb. 'Cannot' would imply physical or logical inability. The distinction matters for tone and authority.
“She does not reveal the A–E mapping. Not to the reviewing advisors, not to me before the round closes, not to anyone outside the schema gate.”
Common mistake: Italian and French speakers often default to 'cannot' for everything, which understates a principled refusal. Peer Review 'does not reveal the mapping' — by design and contract, not because she lacks the technical means.
Em-dash interruption replaced by full stops or commas for in-house voice
This post uses long en-dash interruptions sparingly and uses full stops to break clauses instead. The British editorial register prefers clean sentence boundaries over em-dash parentheticals, especially in instructional writing.
“She publishes the relabelled outputs to the four reviewing advisors. Each advisor sees four outputs, lettered, not their own.”
Common mistake: Learners influenced by AI-generated text overuse em-dashes for emphasis. Replace them with a full stop ('She publishes the relabelled outputs to the four reviewing advisors. Each advisor sees four outputs, lettered, not their own.') for a more authoritative voice.
Anaphoric 'That is...' to label a previous idea
Using 'That is' (or 'That sparseness is...', 'That sharpness is...') at the start of a sentence picks up the previous statement and names it. It is a high-register move that builds rhythm and emphasises the takeaway.
“That sparse consultation graph is intentional.”
Common mistake: Learners try to inline the label ('This sparseness, which is intentional...') and the sentence loses force. Separating it into a standalone 'That sparseness is intentional.' is sharper and more idiomatic.
Comprehension Questions
- 1.What is Peer Review's core job in the Council's Round 2?
- 2.Why does the post argue that anonymisation matters even though the advisors are AI?
- 3.Which of the three review questions does the post call 'the question that earns the Council's fee', and why?
- 4.Name two things Peer Review explicitly refuses to do.
- 5.How might you apply Peer Review's anonymisation principle to a real SME's internal decision protocols, for example a senior team running a strategic review?
Run your own diagnostic
Use the same Strategic Council I run my own decisions through. The assessment preview is free. The specific central human intelligence it is based on is verified in person during the call.
Start the free diagnostic →