# 14 — Hybrid platform vs OEM-dock autonomy

**Added:** 2026-09-15  
**Date:** 2026-09-15  
**Index:** [`CHANGELOG.md`](CHANGELOG.md).  
**Owner:** AT / Velocity EU  
**Audience:** founders.  
**Companion to:** [`11-platform-vs-robotaxi.md`](11-platform-vs-robotaxi.md) (16 August 2026). This note **reinforces** 11’s working answer — build platform **A** first, design the schema for hybrid **C**, do **not** buy docks in seed — against 2026-09-15 competitor reality. It does **not** replace that memo.  
**Also:** [`03-competitors.md`](03-competitors.md), [`13-bulgaria-own-field-experiment.md`](13-bulgaria-own-field-experiment.md), [`briefs/competitors-2026-09-15.md`](briefs/competitors-2026-09-15.md), [`product/option-c-hybrid-fleet.md`](../product/option-c-hybrid-fleet.md).

**Question:** should Canopy evaluate a fully autonomous service like a Tesla robotaxi by **owning docks / fleet**, or by **dispatching jobs onto operator-owned original equipment manufacturer (OEM) autonomy kits** (XAG / DJI / peers)?

No invented total cost of ownership (TCO). Dollars that are not published stay **HYPOTHESIS**.

---

## Three paths (same job object)

| Path | Meaning | Car analogue | Who owns the bird / dock |
|------|---------|--------------|--------------------------|
| **A — Platform** | Match landowners to licensed crews; live map; receipt; cue | Uber | Operators |
| **B — Owned robotaxi** | Canopy fields automated docks + birds and flies jobs | Waymo | Canopy |
| **C — Hybrid** | Platform first; `operator_id` can later be `canopy_fleet` when a week has no crew | Tesla robotaxi ambition (company cars + later owner network) | Mix |

**Working answer (16 August 2026, still current):** build **A**, design schema for **C**, do **not** buy docks in seed. The 2026-09-15 map (Guardian Agriculture dead; XAG XA1 dock launched China-first July 2026, EU dealer path to be determined; Rantizo spray ops sold → AcreConnect software) makes that sequencing *more* important, not less.

---

## What “fully autonomous” actually means in agricultural spray (2026)

OEM autonomy today is mostly:

- **Mission autonomy** under a remote pilot / European Union Aviation Safety Agency (EASA) Specific-category operation — not “no human in the loop.”
- **Dock kits** that auto takeoff, land, charge, sometimes swap payload / mix (XAG Smart Farm 3.0: XA1 dock + LM1 mixer, announced 6 July 2026).
- **Large autonomous aircraft** for big acreage (Pyka Pelican — United States / Brazil estates), wrong shape for Bulgarian plots under 2 ha.
- **Survey docks** (for example DJI Dock + Matrice class) farther along type-approval paths in some countries than **spray** docks.

A dock does **not** delete: labelled product rules, Bulgarian treatment logging, ferry to the plot, wind / rain abort, or **who is the unmanned aircraft system (UAS) operator of record**.

**Implication:** “Tesla robotaxi” for Canopy is **dispatch + utilisation + receipt**, not “delete regulation with software.” Same argument as 11; the new facts do not walk it back.

---

## Who is building the autonomy layer (where they operate)

| Player | Autonomy offer | Where it shows up | Implication for Canopy |
|--------|----------------|-------------------|------------------------|
| **XAG** | Smart Farm dock + mix / refill kit | Asia-Pacific (APAC) strong; European Union (EU) dealers; Western drone-as-a-service (DaaS) fleets | Closest “buy the robotaxi kit” path for multirotor spray |
| **DJI Agriculture** | Agras mission autonomy; Dock products stronger in **survey** | Dominant Bulgaria (BG) / Romania (RO) / Serbia (RS) / Greece (GR) fleets | Bulgaria crews already speak Agras; spray-dock OS still thin |
| **Hylio** | High autonomy / swarm; National Defense Authorization Act (NDAA) US | US + Latin America | Irrelevant to the BG seed; watch for US later |
| **Pyka** | Fully autonomous aircraft, big acre | US, Brazil | Not <2 ha Bulgaria |
| **Guardian Agriculture** | Was electric vertical take-off and landing (eVTOL) spray | US — **shut down 2025** | Warning: capital-heavy autonomy hardware dies |
| **AcreConnect / American Autonomy** (ex-Rantizo software) | Software for human Federal Aviation Administration (FAA) Part 137 crews | US (~30 states claim) | Pattern to copy for **platform**, not docks |
| **FarmFleet** | Marketplace + ops | Ukraine | UX analogue; different regulations |

Full named-contractor map: [`03-competitors.md`](03-competitors.md). **White space remains the job operating system (OS)** — especially EU / BG receipt + live landowner view + small-plot batching — not inventing a better XA1.

---

## Stress test: own docks vs OEM kits on the network

### Option A+ (recommended near term): platform that is autonomy-ready

**Build:** booking, batching, live customer view, Bulgarian Food Safety Agency (BFSA)-shaped receipt, satellite / weather cue, `operator_id` that can later be `canopy_fleet`.  
**Buy / partner:** whatever bird the local crew already flies (Agras / XAG).  
**Evaluate autonomy by:** integrating **operator-owned** dock telemetry when a BG / RO dealer installs an XA1-class kit — Canopy dispatches the *job*, the OEM runs the *steel*.

| Must be true | Falsifier |
|--------------|-----------|
| Crews will accept software dispatch | They ignore the app; WhatsApp only |
| Landowners pay for live + receipt | They only care about €/ha |
| Schema supports `canopy_fleet` later | A hard fork is required = painted into A-only |

This is 11’s punch list, unchanged.

### Option B: Canopy-owned robotaxi docks in seed

| Must be true | Falsifier |
|--------------|-----------|
| Demand exists in a cell with **no willing crew** | Crews exist; Canopy competes with its own supply side |
| Dock TCO < hired crew for that utilisation | Birds idle; ferry kills unit economics |
| Canopy can be legal operator of record | Liability / insurance / Specific ops block |
| Chemistry / mix automation approved for the products in use | LM1-class kit unused; a human still mixes |

**Capital stance:** **do not fund B in seed** unless the Bulgaria own-field experiment plus outside-crew tests show **persistent no-show in a paying cell**. Guardian Agriculture’s shutdown is grade **B** disconfirming evidence for hardware-first autonomy startups.

### Option C: hybrid (Tesla-like), correctly sequenced

1. Prove A on concierge + outside crews — [`13-bulgaria-own-field-experiment.md`](13-bulgaria-own-field-experiment.md).
2. Instrument utilisation gaps (“jobs requested, zero accepts within the window”).
3. Only then pilot **one** dock cell as `canopy_fleet` — preferably **OEM kit capital expenditure (CapEx)**, not a custom airframe.
4. Keep third-party crews as the scale path (Tesla’s unfinished owner-network lesson: company steel alone does not equal network effects).

---

## Decision matrix (what to fund when)

| Signal from Bulgaria / the Balkans | Fund | Do not fund |
|------------------------------------|------|-------------|
| Live + receipt + outside crew works | Platform depth, more crews | Docks |
| Demand with repeated no-shows in one oblast | Single OEM dock pilot as `canopy_fleet` | Multi-oblast Canopy factory |
| Only self-fly works | Operator-tooling pivot or pause the Uber claim | Robotaxi narrative |
| Spray legal path blocked; survey works | Survey marketplace first | Spray docks |

---

## Bottleneck belief (autonomy)

> Canopy can capture Tesla-like upside by owning the **job OS and dispatch**, and by treating OEM docks as **capacity types on the same queue** — without buying steel until a measured crew shortage appears in a paying geography.

If false (OEM docks come bundled with OEM marketplaces that freeze Canopy out), the company must either partner early with XAG / DJI dealer networks or accept analytics / receipt-only positioning. That failure mode is **HYPOTHESIS** — not observed in Bulgaria in this pass.

**Watch list (2026–27):** XAG / DJI dealer marketplace moves in the EU; EASA paths for docked spray; whether AcreConnect-class software expands to the EU.

---

## Recommended evaluation plan (research, not CapEx)

| Week | Action | Output |
|------|--------|--------|
| 1–2 | Interview 2–3 BG / RO Agras or XAG operators about dock interest and no-show patterns | Notes graded B / C |
| 2–4 | Run the Bulgaria own-field experiment ([`13-bulgaria-own-field-experiment.md`](13-bulgaria-own-field-experiment.md)) | B1–B6 results |
| 4–6 | Price one XA1-class kit + insurance + operator-of-record cost as a **HYPOTHESIS TCO** sheet | Go / no-go on even a single dock pilot |
| 6–8 | If no-shows exist *with demand*, design one-cell `canopy_fleet` pilot | Else stay pure A |

---

## Verdict

| Question | Answer |
|----------|--------|
| Should Canopy evaluate a fully autonomous service? | **Yes — as a capacity type and future Option C**, not as seed identity |
| Own docks now? | **No** — unless crew shortage is measured on paying jobs |
| Buy XAG / DJI autonomy later? | **Prefer OEM kits on the network** over custom Canopy steel |
| Closest competitor pattern to copy | AcreConnect-style crew software (US) + FarmFleet-style marketplace UX (Ukraine) + the EU / BG receipt wedge |
| Biggest autonomy risk | Spending like Guardian Agriculture before the OS has liquidity |

11’s five “what we do now” points still stand: handshake crews, `operator_id` ready for `canopy_fleet`, no raise to buy XA1 / DJI Dock / Pelican, watch those kits as **supply we can dispatch**, go / no-go only after 10–30 paying jobs *and* a week that was “no crew.”

---

*Autonomy-path stress test for Canopy research — 2026-09-15.*