JackDaw in the pilots

Crop suitability, frost and growing-degree indicators, farm and market structure for direct sales: the questions the Beta was built for, and the ones it does not answer.

Which farm and adviser questions does JackDaw Beta support, and what did the pilots find limits its use?

Area-based questions: characterising land use, industry and climate, crop suitability, agroclimatic indicators, and the short-food-supply-chain checks (farm scale, block size, LPIS-based suitability for direct sales, local demand). The pilots found the limit in the months of manual work needed to curate local, multilingual data, and in uneven digital skills; it does not decide subsidy eligibility.

Shipped sample prompts: characterise an area; check a crop's suitability
Four short-food-supply-chain tools, including LPIS-based crop suitability for direct sales
Used in the Greek, Finnish, Czech–Bavarian and Spanish pilots (D4.2 Table 7)
D7.3: data curation took months of manual work; JackDaw needed tuning for each language

Open in the interactive book · missions, mini-games and memory cards there

JackDaw for farmers and advisers: what it answers today

main tellingdepth: standard3 min markdown

JackDaw ships with sample queries that show what it was built for, and D4.2 uses them as benchmark prompts in evaluation: characterise this area in terms of land use, industry and climate, and is this area suitable for a given crop? Both start with the farmer or adviser drawing the area on the map.

Questions the Beta can work on

Climate for the field. Sixteen agroclimatic indicators, among them growing degree units, frost occurrence and precipitation patterns, from ERA5-Land history, WorldClim climatology, IPCC RCP 4.5 projections for the 2050s and Open-Meteo short-term forecasts.
Crop suitability. D7.2 shows JackDaw combining land cover, elevation and weather to judge whether a particular plot suits potatoes, where a pure language model gave generic potato advice.
Selling locally. Four short-food-supply-chain tools assess farm scale distribution, agricultural block size, LPIS-based crop suitability for direct sales, and local market demand potential. This links to the food-chain work in short food supply chains.
Policy context. RAG over the PoliRuralPlus policy documents answers with passages from them, not from the model’s memory.

What it does not do

None of the eighteen tools in D4.2 checks CAP payment eligibility, reads a farm’s own records, or issues agronomic prescriptions. A JackDaw answer describes conditions in an area; decisions about payments and treatments stay with paying agencies, advisers and the farmer.

Where the pilots used it

D4.2’s pilot table lists JackDaw in four core pilots: Central Greece (with MAATool, strengthening regional advisory capacity and AKIS cooperation), Mallusjoki (with Raven, for municipal spatial analysis), the Czech–Bavarian border (1 × 1 km attractiveness grids comparing south-western Bohemia and eastern Bavaria), and Spain (mapping the Roman sites of Segobriga, Valeria and Ercavica as one heritage corridor).

What the pilots learned

The policy brief drawn from the nine pilots (D7.3, Annex 10) names the bottleneck: preparing rural, multilingual, domain-specific data for JackDaw and the PoliRuralPlus GPT Advisor took months of manual annotation, formatting and validation. JackDaw needed custom linguistic tuning for diverse EU languages and dialects, and many rural actors lacked the digital skills to contribute. The brief’s answer is institutional, not technical: fund data curation, train digital stewards inside rural institutions, and connect AI tools to AKIS and certified advisers.

For a farmer, the practical rule follows from the evidence trail: use JackDaw to screen an area quickly, then take its tables to an adviser.

Builds on: JackDaw, the PoliRural+ GeoAI agent