Archetype 01 — What's actually deployed

Archetype 01 — What’s actually deployed

A vendor-sweep primer for non-specialist audiences.

Header

What this talk is for

The audience knows AI is in the news but doesn’t have a working picture of what’s actually deployed in food and farming. They’re not hostile, not enthusiastic — they want a map. This talk gives them one and seeds the literacy-as-empowerment stance by ending on the who benefits question.

Run-of-show

1. Opening — what “agrifood AI” actually covers (5 min)

Frame. Agrifood AI is not just robots in fields. It’s the whole value chain — from seeds and inputs through on-farm production, processing, distribution, retail, and waste recovery. Give the audience the seven-cell chain (simplified from v4 taxonomy to a five-cell version for a non-specialist talk: inputs, growing, processing, retail, waste).

Anchor. The European Parliament EPRS study (2023) — scans/2026-07-initial.md — covers this whole chain. Mention as the comprehensive reference, not on screen.

2. Segment one — on-farm production (10 min)

The five biggest names globally. Use as concrete anchors:

Units used: units/john-deere-see-and-spray.md, units/bayer-climate-fieldview.md, units/dji-agriculture-global-export.md, units/xag-china-drone-leader.md, units/lely-astronaut.md, units/yamaha-fazer-agricopter-drone-japan.md, units/spread-techno-farm-vertical-lettuce-japan.md.

Critical move (do not skip). Name what each one claims and what each one measures. Bayer / John Deere claim input reduction (less herbicide, less fertilizer); DJI / XAG claim productivity at scale; Lely claims labour conditions (succession, fatigue, lifestyle); Yamaha claims Japanese-rice precision aerial spraying; Spread claims indoor CEA labour-substitution at scale. These are different claims, not interchangeable.

East-Asia layer — say this if audience will absorb it (otherwise skip). Japan anchors an equipment-vendor industrial automation pattern — Yamaha’s unmanned helicopters and Spread’s vertical farms are mature products from industrial-machinery majors applying their competence to agriculture under an ageing-farmer-labour-shortage imperative. China anchors a state-vendor hybrid with provincial autonomy (DJI, XAG, Alibaba ET Agricultural Brain). Korea anchors a state-anchored cluster programme with vendor participation (Smart Farm Innovation Valley; ioCrops deploying into Japan). The unit units/japan-korea-agrifood-ai-pattern.md is the corpus’s anchor for this East-Asia cluster taxonomy.

3. Segment two — post-farm: processing, retail, waste (8 min)

Move down the value chain. The audience often assumes “agrifood AI” means on-farm; show that it’s now pervasive downstream.

Anchor quote (optional). If audience is sympathetic to producer voices, drop in Jeff Torrie: “If we didn’t invest in new technology, there wasn’t going to be succession. That’s what it came down to.” (quotes/producers/torrie-jeff-lely-succession.md). This is the family-farm-succession motivation, distinct from vendor efficiency framing. Note: 2018 source, flagged historical.

4. Segment three — the data underneath (8 min)

The pivot. Up to now the talk has been about AI. Now: AI runs on data. Whose data, what kind, who controls it?

Three data postures (simplified from v4 taxonomy):

• Open / public — SoilGrids, Copernicus Sentinel, USDA Ag Data Commons. Anyone can use them.
• Proprietary / vendor — Climate FieldView, John Deere Operations Center, AGCO PTx. Vendor controls access.
• Cooperative / commons — JoinData (Netherlands), NAPDC (US federally-funded cooperative-development), Indigenous-led frameworks.

Anchor units: units/open-data-ecosystem.md, units/proprietary-farm-data.md, units/joindata-netherlands.md.

Critical move. Name the dark-data problem — data that is collected but never surfaced for broader use. units/dark-data-agrifood.md. This is the inequality problem in agrifood AI: small farmers and cooperatives generate data; vendors aggregate it; the value flows back to the vendor. The audience should leave knowing this is a structural question, not an abstract one.

5. Segment four — Canada specifically (5 min)

The Canadian anchor. Even for a global talk, the close should land close to home.

• 1.8% vs 12.2% — Canadian agricultural AI adoption (Q2 2025) vs other Canadian industries. Source: Statistics Canada, contextualised by FCC/Deloitte (July 2026). Unit: units/fcc-canada-ai-adoption.md.
• Haven Greens — Canada’s first fully automated AI-powered greenhouse, King City Ontario. Unit: units/haven-greens.md.
• FCC ecosystem-not-technology framing — Canada’s gap is diagnosed by FCC as systemic (fragmented infrastructure, talent, capital, governance) not technological. Unit: units/fcc-ecosystem-not-technology.md.

Anchor quote (optional). If audience responds to producer voices: Jay Willmot (Haven Greens founder) on local demand. quotes/industry-executives/willmot-jay-haven-greens-local-demand.md.

5b. Optional cluster-pattern detour (3-4 min, skip if time pressure)

If the audience is engaged enough for a third-tier observation, the vendor-sweep can pivot to why are vendor deployments organized the way they are?. The canonical answer lives in talks/cluster-pattern-taxonomy.md: six regional cycles have surfaced thirteen named cluster-patterns and three cross-region observations, including:

• Cluster patterns are layered-mix (observed in LAC, MENA, EU-cluster-pattern)
• Cluster-with-three-structures is cross-region pattern (Brazilian seed AI vs. Spanish cooperative AI)
• Cluster-with-state-substrate as substantive pattern (Argentine beef AI SENASA + SIGSA + World Bank financing)

Worth surfacing only if the audience is asking the meta-question. Not a default for the 45-min version.

6. Close — the literacy question (4 min)

The frame. The point of the talk is not “AI is good” or “AI is bad.” The point is that AI in agrifood is real, deployed, and structured by who owns the data. Literacy about that structure is the first step toward any informed position.

Three questions to leave the audience with.

1. When you hear “AI in agriculture,” do you know what part of the value chain is being talked about?
2. Do you know whose data trained the model?
3. Do you know who captures the value from the data once it’s aggregated?

These are not rhetorical. They are the working questions for the rest of the field guide.

Q&A handles

Common audience questions and the units they map to:

• “What about ChatGPT for farmers?” → units/root-ai.md (FCC’s free generative AI extension assistant for Canadian farmers, July 2026). Worth naming that this is the first deployment of a foundation model as a farmer-facing extension tool in Canada.
• “Is China ahead?” → scans/2026-07-china-deepening.md for the export / policy / import signal layers; units/dji-agriculture-global-export.md, units/xag-china-drone-leader.md, units/pinduoduo-smart-agriculture-competition.md for vendor detail.
• “What about MENA?” → scans/2026-07-mena-scan.md for the cluster scan; four anchor units (units/taranis-israel-crop-intelligence.md Israeli venture-funded; units/uae-adafsa-ai-management-certification.md UAE state-as-standards-setter — world’s first ISO 42001:2023 AI management certification in agriculture; units/uae-date-palm-ai-genetic-diversity.md UAE state-as-deployment-of-record + cultural-stewardship; units/lebanon-agrytech-accelerator-agrismart.md Lebanese startup-ecosystem partial-focus); plus Morocco Al Moutmir (OCP Group) and Tunisia RoboCare from cycle-5. MENA cluster is now corpus-distinct as a five-cluster-pattern observation refined from cycle-4’s three-sub-pattern.
• “What about Spanish cooperatives?” → scans/2026-07-spanish-cooperatives-ai.md for the cluster scan; three anchor units (units/spain-cooperative-agrifood-ai-cluster-pattern.md cluster-pattern-with-three-structures observation; units/spain-cooperative-covap-ai-deployment.md first Spanish cooperative-led AI deployment-of-record with 5 named AI tracks; units/mondragon-corporation-cooperative-federation.md Mondragón institutional-federation-anchor substrate since 1956, €11.213B 2024 revenue, 70,000+ workers). Spanish cooperative AI cluster-with-three-structures observation parallels Brazilian seed AI cluster-with-three-structures with structurally-distinct Tier-2 substrates (Mondragón federation-institutional vs. Bayer/Syngenta/BASF/Corteva multinational-corporate-pipelined).
• “What about right to repair?” → vendor lock-in is the bridge; mention as a structural concern. Not the spine of this talk but worth naming if asked.

Freshness check

Before delivering this talk, re-verify the freshness of these anchor units (each carries last-verified: 2026-07):

• units/fcc-canada-ai-adoption.md — annual refresh; re-check Statistics Canada for newer Q2 2026 figures.
• units/loblaw-pcxpress-chatgpt.md — confirm ChatGPT integration still live.
• units/dji-agriculture-global-export.md — confirm deployment figures.
• units/lely-astronaut.md — confirm unit count (was 50,000 across 50 countries).

Substitutions

• If audience is policy / leadership, swap segments 2 and 3 (lead with data sovereignty, then on-farm deployment). This is roughly archetype 03.
• If audience is academic, deepen segment 3 with units/dark-data-agrifood.md, units/farm-data-ownership-critical.md, and the open-source framework unit (units/open-source-in-agrifood-framework.md).
• If audience is farmer co-op, swap to archetype 02 wholesale.

What this archetype is doing in the methodology

This is the entry-point talk — the one a presenter gives when they need to introduce the field. Its job is to give the audience a working vocabulary and a literate posture. It deliberately does not commit to a strong analytical spine (vendor-sweep is the most neutral). Once the audience has the vocabulary, subsequent talks can land harder.