Where AI Value Lands

Ecosystem Business Archetypes (Choudary)

5 min read

Ecosystem Business Archetypes (Choudary)

Sangeet Paul Choudary’s four-archetype framework for horizontal business models that emerge once a vertically integrated value chain unbundles into a layered ecosystem. The vocabulary predates the AI wave (the canonical statement is the 2022 teardown — see choudary-ecosystem-teardown) but it is the cleanest off-the-shelf taxonomy for naming where in the ecosystem a firm sits and how that position captures value.

The four archetypes are not stack layers (Axis 1) and not levels of autonomy (Axis 2). They are positions in a horizontal ecosystem — orthogonal cuts on the value chain.

The four archetypes

ArchetypeFacesMechanismCanonical examples
AggregatorConsumer (B2C)Aggregates demand; owns consumer engagement and data; gatekeeps producer market accessFacebook, Amazon, Netflix, Apple, Google
IntegratorB2B (producer ↔ distributor)API switchboard: integrates supply-side product APIs with demand-side distribution surfacesAmadeus, Sabre (travel); Plaid, Galileo (BaaS)
InfrastructureWhole ecosystemProvides core production substrate, standards, roadmap; coordinates ecosystem firms toward a shared outputiOS, AWS, Shopify, Autodesk Construction Cloud, DeepMind
CapabilityEcosystem firms (embeddable)Specialised function licensed via APIs / modules; embedded inside aggregators, integrators, infrastructuresStripe, identity providers, healthcare diagnostic AI

Aggregator

Consumer-facing. Three functions: (1) provisioning consumer services, (2) capturing and monetising consumer data, (3) matchmaking producers ↔ consumers. The strategic asset is ownership of consumer engagement — increasingly scarce, increasingly valuable as a gatekeeping position. Aggregation Theory’s home archetype.

Integrator

B2B switchboard between producers and distributors. Three functions: (1) production-side service integration (one-stop storefront for distributors), (2) data exchange between sides, (3) consumer analytics fed back to producers. Quietly dominant in industries that look fragmented from the outside (travel inventory, payment rails, banking-as-a-service).

Infrastructure

The substrate the ecosystem runs on. Three functions: (1) publish a future vision and roadmap producers can align with, (2) provide knowledge services (data, ML models) to ecosystem actors, (3) attract third-party service providers (AWS Marketplace, Shopify Fulfilment Network). Captures value by being the coordination layer around which others organise.

Capability

Specialised module licensed via APIs. Three functions: (1) provisioning a non-commoditised capability (requires IP / data advantage), (2) architecting for ease of embedding (developer experience is competitive), (3) using capability-usage data to improve the capability and provide analytics. Survives by out-specialising generalists; loses if commoditised.

Mapping to this project

Choudary’s archetypes sit on Axis 3 — Business Archetype (07_analytical-vocabulary). They give a clean four-way decomposition that complements the project’s fifteen-archetype roster:

ChoudaryThis project (partial mapping)
AggregatorConsumer AI app; AI marketplace / platform
IntegratorSystem integrator; AI agency (when API-bundling); BaaS-style platforms
InfrastructureHyperscaler AI services; Foundation model lab (in part); AI developer tools (Cursor, when treated as developer platform)
CapabilityData services; deep tech startup (when API-licensed); identity / payments / verticalised AI APIs

The mapping is not bijective. OpenAI is Foundation + Capability + (consumer) Aggregator simultaneously. The point of Choudary’s framework is to force a primary-archetype choice and surface mistakes — see cross-archetype-confusion.

Why this matters for “where AI value lands”

The 2022 framework predicts the modern AI wave’s value-capture pattern with little adjustment. Distribution-moat firms (distribution-moat: Cursor, Perplexity, Windsurf) are Aggregators in B2B clothing — consumer-of-developer engagement, data flywheel, gatekeeping the model providers below. Foundation-model labs occupy the Infrastructure archetype (publish a roadmap, provide knowledge services as licensed weights, attract third-party ecosystems). Most failed mid-stack AI startups attempted Capability positions without the IP or distribution to survive commoditisation.

This is also the cleanest answer to the middle-layer-defensibility question: there is no monolithic “middle.” There is an Integrator middle (which can be defensible — Stripe-class switchboards) and an undifferentiated “thin wrapper” middle (which is not an archetype at all in Choudary’s taxonomy — it is a failed Capability).

Relationship to H2_u-curve-of-value

Choudary’s “horizontal business models” thesis is a structural precursor to the U-curve. His claim: as vertical chains unbundle (see vertical-to-horizontal-unbundling), value concentrates in horizontal positions that span the chain. H2 is the AI-era specialisation: among Choudary’s four, Aggregators and Infrastructures capture most surplus (top and bottom of the U), Integrators are bimodal (defensible if switchboard-essential, squeezed if generic), and Capabilities concentrate or commoditise depending on IP defensibility.

Open questions

  • Where does an Integrator’s value sit on the U-curve? Plaid and Stripe both look like “top of U” by economics but sit in the middle by Choudary’s position. → contradiction with H2 or refinement?
  • Is the four-archetype taxonomy stable for AI-era ecosystems, or does agentic infrastructure require a fifth archetype (e.g., Orchestrator — the multi-agent coordination layer described in vertical-ai-orchestration)?
  • How do the archetypes interact with divergent-value-stack-optima? Choudary’s vocabulary is US-stack-shaped; China’s open-weight commoditisation strategy and the EU’s compliance-layer play do not map cleanly onto Aggregator / Integrator / Infrastructure.

Companion taxonomy — Weber’s four AI-startup patterns

Choudary’s framework is a horizontal-ecosystem taxonomy (where does the firm sit in the unbundled value chain?). Weber et al.’s parallel framework ai-startup-business-archetypes-weber is a business-model taxonomy (what does the firm sell, to whom, how does it create value?). Both are four-archetype taxonomies but they answer different questions. Read together they make up Axis 3 of 07_analytical-vocabulary more completely than either does alone:

AspectChoudary (horizontal ecosystem)Weber (AI-startup BM)
LensWhere in the unbundled chainWhat kind of AI business
Question”Position?""Sells what, to whom, how?”
TimePre-AI structuralAI-era empirical (100 Crunchbase startups)
ArchetypesAggregator / Integrator / Infrastructure / CapabilityAI-charged P/S Provider / AI Development Facilitator / Data Analytics Provider / Deep Tech Researcher

Cross-mappings are not 1:1 — e.g., a Choudary Capability that happens to use AI could be a Weber AI-charged P/S Provider or a Deep Tech Researcher depending on whether the AI is productized.

Graph View

Backlinks