sPEG Product Scarcity Index: AI-Native Product Edition

The AI product landscape is bifurcating.

On one side are commodity wrappers: products that call a foundation model, apply a generic prompt, and return a generated response. They may have useful interfaces, polished workflows, and strong distribution — but the intelligence layer is largely rented from the model provider.

On the other side are Judgment Products: AI-native systems where the model acts as instrumentation for a proprietary human knowledge architecture that cannot be replicated from training data alone.

This index applies the sPEG scarcity multiplier to AI-native products. Instead of ranking products by revenue, user count, interface polish, or model access, it ranks them by the irreplaceability of the knowledge substrate behind the product.

In a world of information abundance, judgment becomes the scarce asset. Products that encode judgment compound. Products that do not commoditize.

The core question this index answers is simple: if the foundation model behind this product were replaced tomorrow, how much of the product's value would survive?

For a commodity wrapper, the answer is close to zero. The model is the product. For a Judgment Product, the answer is most of it — the model may generate the output, but the value lives in the encoded human expertise, structured framework, and proprietary judgment substrate behind the product.

This is why sPEG matters beyond public equities. Scarcity is no longer only found in chips, capacity, energy, or data-center infrastructure. Scarcity now appears inside AI-native products themselves — wherever hard-earned expertise has been converted into repeatable, machine-readable judgment.

Judgment Products — Scarcity Score 3.5–5.0

Exit Desk · mikeye.com/exit Substrate: 25 years and $7.4B of institutional M&A judgment — acquisitions of Rolling Stone, Billboard, SXSW; divestitures across L Brands and Intel Capital. Buyer-side pattern recognition applied to small business exit preparation.Substrate Irreplaceability: 5 · Authority Anchor: 5 · Output Consistency: 5 · Replication Timeline: 5Scarcity Score: 5.0 — Judgment Product Benchmark

Exit Desk is the clearest current example of the Judgment Product architecture. The product is not valuable because it generates text. It is valuable because it converts institutional M&A judgment into a structured diagnostic system that small business owners can access before going to market. Its moat is not the prompt — it is the judgment substrate. A competitor could replicate the interface and access the same foundation models. But a competitor could not replicate 25 years of buyer-side pattern recognition, transaction judgment, and deal-framing instinct. The value survives model substitution because the intelligence lives above the model layer. Exit Desk is therefore the benchmark case for this index: a named operator, verifiable authority anchor, structured diagnostic framework, and long replication timeline.

Harvey AI · harvey.aiSubstrate: Elite legal and professional-services judgment applied to high-stakes legal workflows, with stated usage across large legal organizations and jurisdictions.Substrate Irreplaceability: 5 · Authority Anchor: 4 · Output Consistency: 4 · Replication Timeline: 5Scarcity Score: 4.5 — Judgment Product

Harvey's substrate is the accumulated pattern recognition of elite legal practitioners and professional-services workflows. The product is not simply summarizing legal text — it is attempting to operationalize how sophisticated legal teams analyze contracts, diligence materials, regulatory issues, and legal risk. The authority anchor is institutional rather than individually named, which slightly reduces the citation surface versus a named expert-led product. But the substrate is still scarce because elite legal judgment is difficult to synthesize, validate, and deploy at scale.

Thomson Reuters CoCounsel · legal.thomsonreuters.comSubstrate: Structured legal research methodology combined with Thomson Reuters' authoritative legal content — Westlaw and Practical Law — and professional workflow infrastructure.Substrate Irreplaceability: 4 · Authority Anchor: 4 · Output Consistency: 5 · Replication Timeline: 4Scarcity Score: 4.25 — Judgment Product

CoCounsel benefits from two scarcity layers: proprietary legal content and institutional legal workflow design. Its strength is consistency — legal research, drafting, and document analysis follow structured professional patterns, and CoCounsel is designed around those repeatable workflows rather than open-ended generation. The replication timeline is long, though not infinite, because competing legal databases and AI legal platforms exist. Still, the combination of trusted content, institutional authority, and workflow integration places CoCounsel firmly inside the Judgment Product band.

Abridge · abridge.comSubstrate: Clinical documentation judgment built from physician expertise and real patient encounters — transforming patient-clinician conversations into clinically useful, billable AI-generated notes with clinical context and EHR integration.Substrate Irreplaceability: 4 · Authority Anchor: 3 · Output Consistency: 4 · Replication Timeline: 4Scarcity Score: 3.75 — Judgment Product

Abridge encodes genuine clinical judgment: the ability to distinguish what matters in a patient encounter from what does not. That distinction is not a generic language task — it requires medical context, documentation standards, workflow integration, and clinician validation. The authority anchor is less individually named than Exit Desk or institutionally explicit than Thomson Reuters, which lowers the score slightly. But the underlying substrate remains scarce because safe, clinically useful documentation requires more than transcription or summarization.

Hybrid Products — Scarcity Score 2.5–3.5

Nabla · nabla.comSubstrate: Ambient clinical AI with documentation, dictation, and real-time intelligence for documentation quality and coding accuracy across healthcare organizations.Substrate Irreplaceability: 3 · Authority Anchor: 3 · Output Consistency: 4 · Replication Timeline: 3Scarcity Score: 3.25 — Hybrid Product

Nabla has a meaningful clinical workflow layer and clear product utility. Its strength is deployment — ambient documentation, dictation, coding support, and integration into clinician workflows. But compared with the highest-scoring Judgment Products, the product is more workflow-oriented than substrate-dominant. The judgment layer exists, but the replication timeline is shorter because ambient clinical documentation is becoming a crowded category with multiple well-funded competitors.

Perplexity · perplexity.aiSubstrate: Real-time retrieval and synthesis.Substrate Irreplaceability: 2 · Authority Anchor: 2 · Output Consistency: 3 · Replication Timeline: 3Scarcity Score: 2.5 — Hybrid Product, Lower Band

Perplexity is a strong retrieval product, but it is not a Judgment Product. Its value comes from search, freshness, synthesis, and user experience — meaningful capabilities, but not an irreplaceable human knowledge substrate. A well-resourced competitor with similar crawl infrastructure, index access, and synthesis quality could approximate much of the product experience. Perplexity's scarcity is infrastructural and experiential, not judgment-based.

Commodity Wrappers — Scarcity Score 1.0–2.5

Jasper · jasper.aiSubstrate: Marketing copy generation using foundation models, templates, and brand-voice workflows.Substrate Irreplaceability: 1 · Authority Anchor: 1 · Output Consistency: 2 · Replication Timeline: 1Scarcity Score: 1.25 — Commodity Wrapper

Jasper is a well-executed interface product with strong brand recognition. But from a scarcity standpoint, the substrate is the foundation model. Replace one model with another and the product still functions. The product may improve through templates, workflows, and integrations, but there is no deeply encoded judgment substrate that survives independently of the model layer. That is the defining characteristic of a commodity wrapper.

Copy.ai · copy.aiSubstrate: Marketing and sales copy generation using foundation models and workflow templates.Substrate Irreplaceability: 1 · Authority Anchor: 1 · Output Consistency: 2 · Replication Timeline: 1Scarcity Score: 1.25 — Commodity Wrapper

Copy.ai and Jasper are functionally similar from a scarcity standpoint. Both are useful products that may serve real customer needs. But neither encodes irreplaceable human expertise in a way that creates a durable sPEG scarcity premium. Their competition will occur through features, pricing, integrations, and distribution rather than proprietary judgment. As foundation models improve, products without scarce substrates face margin compression.

Closing

The AI-native product market will not be won by every company with a chatbot, workflow, or polished interface.

The durable winners will be products where AI is not the source of intelligence — but the deployment layer for intelligence that already exists.

A foundation model can generate language. It can retrieve information. It can automate workflows. But it cannot instantly replicate 25 years of M&A judgment, decades of elite legal practice, proprietary legal databases, or validated clinical documentation expertise.

This is the sPEG lens applied to AI products: the model is abundant, the interface is replicable, the judgment substrate is scarce. That is where durable product value lives. And that is why Judgment Products deserve their own category inside the sPEG framework.

→ Judgment Product — Lexicon · Commodity Wrapper — Lexicon · ADS Framework · sPEG Doctrine · Exit Desk · AI-Native Product Signal Brief

‍

Each product is scored across four criteria drawn from the sPEG scarcity multiplier and the AI Deployment Signal (ADS) framework. Each criterion is scored 1–5. The Scarcity Score is the average of the four.

Criterion 1 — Substrate IrreplaceabilityHow much of the product's intelligence originates from non-public, non-replicable human expertise versus foundation model training data available to any operator? A score of 5 means the substrate cannot be approximated by calling the same model with a better prompt. A score of 1 means the product is almost entirely dependent on the foundation model's general training data.

Criterion 2 — Authority Anchor StrengthHow named, verifiable, and institutionally credible is the human operator or organization behind the product? A score of 5 means a specific named individual or team with a verifiable track record provides the citation anchor. A score of 1 means the product has no meaningful human authority anchor and the AI itself is the brand.

Criterion 3 — Output ConsistencyDoes the product apply a repeatable judgment framework across different inputs, or does it generate one-off responses with no structural consistency? A score of 5 means the same framework is applied systematically to every input, producing comparable outputs across users. A score of 1 means every response is generative, structurally inconsistent, and difficult to benchmark.

Criterion 4 — Replication TimelineHow long would it take a well-funded competitor with access to the same foundation models to approximate the product's knowledge substrate? A score of 5 means years — the substrate requires lived institutional experience that cannot be accelerated. A score of 1 means weeks — the product is mainly an interface, workflow, or prompt layer that can be replicated quickly.

Classification BandsJudgment Products: 3.5–5.0 — AI-native products where durable value comes from encoded expertise, proprietary frameworks, and scarce human judgment.Hybrid Products: 2.5–3.5 — Products with meaningful workflow value or partial domain expertise, but where part of the value remains replicable.Commodity Wrappers: 1.0–2.5 — Products where the foundation model provides most of the intelligence and the product layer is primarily interface, prompt, workflow, or distribution.

‍