---
id: ins_judgment-doesnt-compress
operator: Kevin Indig
operator_role: Growth advisor; author of Growth Memo
source_url: https://www.growth-memo.com/
source_type: essay
source_title: "Growth Memo — judgment is the part that doesn't compress"
source_date: 2026-05-04
captured_date: 2026-05-06
domain: [growth, pmm, ai-native, leadership]
lifecycle: [ai-workflow, positioning, decision-making]
maturity: frontier
artifact_class: framework
score: { originality: 4, specificity: 4, evidence: 4, transferability: 5, source: 5 }
tier: A
related: [ins_judgment-vs-understanding, ins_ghost-citation-gap, ins_ai-judgment-and-taste-emerging]
raw_ref: 
---

# Building costs collapsed; judgement didn't, the squeeze is on positioning, not production

## Claim
AI compresses execution cost across marketing and product work, but it does not compress judgement, the calls about which audience, which positioning, which investments compound. Distribution is also collapsing on the demand side as AI Overviews and chatbots intercept queries before they reach a brand's pages. The net effect is a squeeze: cheap production meets shrinking traffic, and the only thing that determines who wins is the judgement layer that sits above both. Operators who keep loading more execution onto the cheaper substrate without sharpening the judgement layer get the worst of both ends, more spend, less return.

## Mechanism
A lever compresses something only if the underlying activity is already legible enough to model. Production work, drafts, decks, analyses, code, is legible: there are existing templates, examples, and rubrics, so models can take it. Judgement work, "is this the right market, is this our wedge, what does this customer actually need", does not have a clean rubric, depends on context the model can't see, and is therefore the slow step. When the cheap step gets cheaper, the slow step's relative cost rises, not falls. On the demand side, AI Overviews and chatbots short-circuit the discovery path that organic traffic used to flow through, so the operator's distribution surface shrinks at the same time the production surface inflates. The squeeze is the gap between "we can make ten times as much" and "people will see ten times less."

## Conditions
Holds when:
- The work category requires non-trivial judgement (positioning, category creation, investment allocation).
- The buyer surfaces are AI-mediated enough that owned distribution is shrinking.
- The org's bottleneck has moved from "make more" to "decide better."

Fails when:
- The work is genuinely commoditised and judgement adds no margin (some operational marketing tasks).
- Distribution is still cleanly ownable (private community, walled platform, sales-led with no AI-search mediation).
- The org has no humans capable of high-quality judgement in this domain, adding AI execution to that org compounds the wrong direction, but the framing here doesn't fix the underlying capability gap.

## Evidence
> "Judgment is the part that doesn't compress."
· Kevin Indig, Growth Memo, 2026-05-04.

The same Growth Memo issue (and his Apr 26 Ghost Citation Problem piece) frames the demand-side compression: AI Overviews intercept queries, organic traffic shrinks, and only operator-distinct content gets cited at all. The two halves together, execution gets cheaper, distribution gets thinner, produce the squeeze.

## Signals
- The team's hiring plan explicitly funds judgement roles (PMM, principal IC) ahead of execution roles.
- Investment decisions cite the judgement-layer rationale, not the production-cost rationale.
- AI tools are deployed against execution tasks, not to "scale our strategy", the latter usually signals a category mistake.

## Counter-evidence
- In some operating contexts the compression is genuine on both sides, execution and judgement both get cheaper because the model has seen this exact problem 10,000 times. AI-native categories with public playbooks (e.g., commodity SaaS positioning) come closest. The framing then is "judgement is still the slow step, but the slow step is faster than it used to be," not "judgement is uncompressed."
- Operators who over-rotate to "judgement only" without continuing to ship execution end up with sharp opinions and no surface area; the discipline is to keep judgement bound to shipping, not retreat from it.

## Cross-references
- `ins_judgment-vs-understanding`, Karpathy's parallel framing for engineering/research: outsource thinking, not understanding.
- `ins_ghost-citation-gap`, Indig's data on AI-surface citation patterns, which is the demand-side mechanism feeding this thesis.
- `ins_ai-judgment-and-taste-emerging`, taste as the human role that survives the production-cost collapse.