Taker / Shaper / Maker
A strategic-adoption typology from wef-ai-in-action-2025 (echoed in non-tech-digital-core-synthesis). Names three positions a non-tech enterprise can take relative to foundation models.
The three archetypes
- Taker — integrate commercial off-the-shelf AI into existing workflows. Minimal customization. The model is rented as a service; the firm’s job is to wire it into business processes.
- Shaper — augment an existing foundation model with proprietary data, fine-tunes, RAG over internal knowledge, or domain-specific evaluation. The base model is rented; the differentiation is in the layer above it.
- Maker — train bespoke foundation models from scratch. Full vertical ownership of model weights, training data, training compute. Only a small fraction of non-tech firms have the capital or talent to do this.
How it relates to the other ladders
The taker/shaper/maker axis is a strategic-positioning axis, distinct from but interacting with:
| Axis | What it answers |
|---|---|
| taker-shaper-maker | How much of the model do we own / customize? |
| H1_L0-L7-ladder | How much of the substrate stack do we own? |
| enterprise-adoption-ladder | How deeply has AI penetrated our operations? |
| digital-core | Have we built the architectural preconditions? |
A firm can sit anywhere on the taker/shaper/maker spectrum at any phase of the adoption ladder. The interesting empirical fact is that most non-tech firms are takers, and the bull-managerial frame implicitly assumes this — its case studies (Accenture / SAP / Databricks customers) are nearly all takers and shapers, almost never makers.
What the typology reveals
- The strategic question is not “should we use AI?” but “what is our position relative to the foundation model?” This is more useful than the binary “AI vs. no AI” framing common in earlier consultancy material.
- Most differentiation lives at the shaper layer. Taker has no moat (everyone rents the same API). Maker is gated by capex. Shaper is where domain knowledge, proprietary data, and workflow integration compound — and is plausibly where the largest population of non-tech firms can build defensible positions.
- Shaper-vs-taker is the consultancy upsell axis. Accenture’s commercial interest is precisely to move clients from taker (cheap, commoditized) to shaper (custom integration work, ongoing engagement). Read the typology with this in mind.
Mapping to the OECD SME taxonomy
oecd-sme-ai-adoption-2025 proposes a buyer-side SME four-stage taxonomy (oecd-sme-adopter-taxonomy) — Novices, Optimisers, Explorers, Champions — that maps approximately onto Taker / Shaper / Maker (see also sme-policy-pathway-novice-to-champion):
- AI Novices → Takers. Renting turnkey gen-AI via cloud APIs; no internal AI capability. The Novice rents narrowly.
- AI Optimisers → Takers transitioning to Shapers. Embedding off-the-shelf AI widely across functions; co-funding training; beginning to restructure operations around it.
- AI Explorers → Shapers. Developing bespoke tools, custom agents, fine-tunes on proprietary data in narrow scope; consuming sovereign compute and curated datasets; investing in long-term R&D. The model is rented; the layer above it is proprietary.
- AI Champions → deep Shapers, occasionally small-scale Makers (e.g. SMEs operating near the frontier — such as those training proprietary ML stacks for rare-disease drug discovery, where the data + model architecture is genuinely the business). Integrated into innovation ecosystems; minimal direct subsidy needs.
Two implications follow:
- The OECD case-study evidence supports the commercial-interest reading. Most SMEs are pure Takers (Novices + Optimisers); the consultancy / policy upsell is precisely to move them into the Shaper layer (Explorers / Champions).
- Empirical support for “most differentiation lives at the shaper layer.” Calvino & Fontanelli (2024) show firms developing AI internally capture significantly larger productivity premia than firms sourcing externally (ai-productivity-firm-level).
Policy instruments differ at each stage: Novices need awareness; Optimisers need skills and pre-approved tools; Explorers need compute and data; Champions need stable ecosystem and frontier infrastructure. This is the buyer-side policy analogue of the supplier-side rent/customize/train framing.
What it hides
- Renting the maker layer. A “shaper” who builds on top of OpenAI is still paying surplus to OpenAI. Taker/shaper/maker says nothing about whether the firm captures or transfers value at the underlying layer. See H1_L0-L7-ladder.
- Distribution. A Maker that nobody uses captures no value. A Taker with a dominant user relationship may capture more than a Maker without one. See distribution-moat.
- The middle dies. The Shaper position assumes the middle of the stack is defensible. middle-layer-defensibility argues this is true only for distribution-embedded platforms (Cursor, Windsurf) — a generic shaper without distribution is a thin wrapper.
- Bloc divergence. In China, the Maker layer is partially commoditized via open-weight-asymmetry, collapsing the cost gap between Shaper and Maker. In the EU, sovereign-ai is an attempt to push national champions toward Maker status via subsidy, which is structurally different from US private-capital Maker plays.
Why the framing persists
It is legible to executive buyers: three boxes, one per strategic appetite for risk and capital. It is also flattering to non-tech firms: it presents Shaper as a credible position rather than admitting that most firms will remain pure Takers. And it is commercially aligned with consultancies: every transition between archetypes is a multi-year engagement.
Use in the paper
Likely useful as a foil: introduce the typology as the consensus consultancy framing, then show what it leaves out — substrate ownership (H1_L0-L7-ladder), distribution (distribution-moat), surplus capture (capital-labor-divergence), and bloc structure (divergent-value-stack-optima). The paper’s value is to surface the dimensions the typology collapses.
Related
- wef-ai-in-action-2025 — primary source.
- non-tech-digital-core-synthesis — restatement.
- digital-core — architectural precondition the typology assumes.
- foundational-enablers — the enabler frame the typology sits inside.
- H1_L0-L7-ladder — substrate-ownership axis the typology collapses.
- enterprise-adoption-ladder — buyer-side maturity ladder; orthogonal axis.
- distribution-moat — strategic asset the typology ignores.
- middle-layer-defensibility — the load-bearing question for Shapers.
- ai-startup-business-archetypes-weber — seller-side companion taxonomy; together Taker/Shaper/Maker (buyer-side) + Weber’s four patterns (seller-side) + ecosystem-business-archetypes (ecosystem-position) form the three-lens Axis 3.
- open-weight-asymmetry — what happens to the Maker layer in China.
- sovereign-ai — EU push toward national-champion Makers.
- oecd-sme-ai-adoption-2025 — SME-side mapping.
- oecd-sme-adopter-taxonomy — the buyer-side decomposition that maps onto Taker/Shaper/Maker.
- sme-policy-pathway-novice-to-champion — staged policy instruments along the same axis.
- sme-ai-adoption-gap — empirical context for the Taker-population size.
- ai-productivity-firm-level — evidence for the Shaper premium.
- H5_ai-as-operational-not-product — applies the Taker/Shaper/Maker decomposition at firm level: the operator-builder is a Taker of AI capability, a Shaper of harness, and a Maker of a non-AI deliverable.