Elzeard, the Market Farmer’s Companion

Planning and monitoring crops is a real headache for vegetable producers, who are more and more numerous to settle to meet the demand for fresh, local and sustainable products. No tool offers them access to standard cultural practices in organic or reasoned agriculture and share practices to get help.

Elzeard wishes to support these producers by developing the Farmer Market’s Companion, a digital platform for optimizing their crop itineraries and personalizing them depending on various constraints such as the soil and climate factors. This digital solution relies on a knowledge base of agroecological market gardening practices, the Knowledge Greenhouse,  which will be fed by the community of producers, agricultural advisers and researchers as a common good. 

If scaled globally, it will help producers to gain yield productivity and hence better income. It will also relieve their daily mental workload and guide them to adapt their practices to become more sustainable.

About 823 000 farms cultivated fresh vegetables across the EU in 2016, mostly small and family farms as in the rest of the world (Eurostat,2017). They currently face a rising number of vulnerabilities: economic, regulatory, environmental or social (isolation, skills, overload, etc.). These mutations have increased pressure on these farmers to keep up with the requirements of efficient practices and the ability to provide safe and high quality vegetables (FAO,2014). The proposed solution is to transition actual agricultural methods to agroecological ones as they could "double food production in entire regions within 10 years while mitigating climate change and alleviating rural poverty" (UN,2011). However, these innovative agroecological practices increase the complexity of crop management, the workload and mental load of farmers (Morel & Léger,2015; Dumont & Baret,2017). Indeed adopting agroecological practices means a wider cultivated vegetable diversity not only to better manage risks in a fluctuating climatic and economic context, but also to fight against pests/diseases to eliminate use of pesticids and reduce carbone print, both by diversity in time (rotation) and in space (plot mosaic) (Isbell &al.,2017; Paut & al.,2019). It thus requires more knowledge, more acceptance of risk, and new methods from already weakened farmers.

The solution carried by Elzeard is organized around two applications. Firstly, The Market Farmer’s Companion is a web and mobile application with three modules:

1. Planning (cultural calendars and rotation plans optimized according to local soil and climate constraints and production objectives);

2. Cultural monitoring (interventions and crops, team planning, observations, managing hazards and suggesting alternative non-chemical treatments); 

3. Diffusion (communicating about harvests and cultivation methods towards the consumption chains, automatically generating administrative and regulatory documents).

Secondly, the Knowledge Greenhouse is an open web portal to access and share knowledge about sustainable market gardening methods and all other resources useful to farmers (stakeholders' map, events, social network, etc.) in order to promote the transition of agrosystems to more environmentally friendly production. It is fed by the community of producers, agricultural advisers and researchers and promoted as a « common good ». 

The data shared via the Knowledge Greenhouse is aggregated and structured to feed the AI algorithms developed for the Companion’s recommander system. Conversely, the field data shared from the Companion enriches the Knowledge Greenhouse's graph network to be analyzed, enriched and reused. The two applications form a virtuous loop of learning operational practices, adjustable to the context of farms.

Vegetables are grown in nearly 200 countries around the world and most often in one of the 500 million small family farms (FAO, 2014). Our target population is firstly dedicated to diversified market gardening which represents a volume of more than 800,000 producers in Europe (Eurostat, 2017) et 75,000 in USA (USDA, 2017). We can classify these producers into three groups per cultivated area: up to 4 Ha, from 4 to 10 Ha and more than 10 Ha. This latter population is more akin to industrial production in field crops and is not part of our initial target. 

Elzeard is committed to a co-construction approach since its beginning. We carried out 10 workshops in 8 various French regions to collect needs from producers, validate their appetite for a digital application and the desire to share data with their peers. We conducted two studies to better understand the sociological profile of small to medium-size farmers and their current use of digital technology (Elzeard, 2019). We regularly interview market gardeners, agricultural advisers, local authorities, research and teaching centers, associations. We identified how our application could unburden them by helping adapting their practices without losing income or adding mental stress.

Using agroecological methods to make food more sustainable is a question of reconciling high productivity by exploiting natural ecosystems while minimizing the impact on the environment. It requires the transfer and sharing of actionable knowledge between researchers, agronomists including farmers, especially those from small family farms like many vegetable producers (Eurostat, 2017). The use of digital technologies, especially in artificial intelligence, can help market gardeners to better understand their agrosystem to optimize planning, rotations and cultivation operations (Pillaud, 2015). Keeping intervention data provides valuable assistance in analyzing the results compared to harvest forecasts to create virtuous learning loops.

No similar application fully dedicated to market gardening sector exists today. Most vegetable growers rather use paper notebooks, calendars, “Velleda” whiteboards or Excel files (for the most advanced ones) to plan and keep track of their activity. But the major alternative has been mainly not to do!! The direct competitors identified to date are American: Crop Planning Software (open-source but unmaintained anymore), and more recently AgSquared and Tend. Those applications lack the crop planning suggestion engine, the exploitation of data and knowledge sharing, limiting their potential impact.

The design of our applications relies on the modelling of the standard technical cultivation itineraries for each crop species: what to plant, when to plant, what associated interventions, how much workload, what rotation... We use Semantic Web technologies to model, aggregate and interconnect all the data needed for each growing vegetable. This semantic data is then exploited by logical reasoning algorithms combined with machine learning techniques, forming the foundation of an innovative service to support vegetable planning service, adaptive and customizable. The challenge comes down to setting up the most suitable usecase scenarios, closest to the daily life of the producer and its constraints (soil, climate...) with intuitive interfaces, by pre-filling the reference data. The major objective is to bring added value for the producer: saving time and productivity, comfort and well-being on a daily basis in its plots. This innovation is carried out with producers by including them in the process of ideation and prototyping of applications through workshops.

To create a vegetable production support application, we use artificial intelligence (AI) technologies. Connectionist techniques (deep learning) require large volumes of data that hardly exist in agriculture due to the impossibility of comparing data from different farms and the lack of source data. Hybridization with symbolic (logical) AI would bootstrap machine learning on smaller volumes of data. We therefore decided to exploit Semantic Web techniques, based on the use of open and interoperable models, semantic graphs and reasoning engines, allowing from:

• Capturing and representing cultural processes and knowledge

• Integrating decision support services

• Mobilizing reference vocabularies (RDF, OWL, SKOS) for data interoperability 

The semantic approach (structuring and enriching data based on controlled vocabularies/ontologies, querying and exploitating linked data networks) is used to construct the knowledge graph that will represent the culture itineraries for the recommendation service of rotation planning. This knowledge graph will interconnect various datasets (agricultural, agronomic, meteorological data, etc.) to model varieties, soil and climatic conditions, cultural operations, sowing / planting / harvesting dates, fertilization optimization, pest control techniques, irrigation, rotation, etc.

Moreover, the use of an open knowledge model, based on common repositories and interoperable formats recommended by the W3C (RDF, OWL, SPARQL) will facilitate the sharing of data in the Knowledge Greenhouse as a common good according to FAIR principles. It will also ease the interconnection with any partner tool, as the sharing of production information up to the final consumer. They will also encourage the development of micro-services with other agricultural service providers (weather, IoT).

The Semantic Web techniques (Gandon, 2018) are used for information sharing, automatic processing of information resource bases and recommendation and reconciliation services. Little or not used in the agricultural field, these technological bricks are today recognized as formidable means of development of new services around the data. They are configurable to be adapted to the domain of use.

Some agronomic resources are available in the interoperable formats of the Semantic Web, recommended by the W3C (RDF, OWL, SPARQL). For example, the AgroPortal website publishes ontologies for creating knowledge bases such as GECO (aggressors, diseases, auxiliaries and their indications in the culture system) or thesauri and vocabularies such as TaxRef, AgroVoc, Crop Ontology, Plant Ontology or FrenchCropUsage…

However, few datasets or models describe the vegetable production or even the species from an agricultural point of view. That is why Elzeard builds an open model of semantic representation of the management of plots and vegetable growing routes with the help of the domain experts, the farmers and the agricultural technical advisors, partners of the project. A PhD student will work to hybrid this model including reasoning axioms with machine learning techniques to optimize intelligence for cultural operations planning. This hybridization of symbolic and connectionist methods are a recent and very interesting topic of research (Ristoski, 2016; Sun, 2019).

Elzeard offers a digital platform to help simplify the monitoring of vegetable farms that wish to apply the agroecological pillars that are the spatial and temporal diversification of cultivated plants. This includes a web app for better understanding and managing the complexity of diversified systems combined to an open web portal for sharing knowledge and experiments about agroecological practices in vegetable production. 

Our immediate goals are to lower the mental load of producers and to improve their working conditions. By promoting agroecological practices allowing better management of fertility and plant health through diversification, Elzeard intends to participate in greater farm autonomy by lowering input costs and optimizing production, making it possible to free up better income.

These economic short-term outcomes link to our long-terme outcomes achieving both environmental and territorial impact. Firstly, by supporting the design of agroecological vegetable farms, The Companion application can help market gardeners to reduce the use of fertilizers and phytosanitary products while increasing biodiversity. The integration of green manures and intermediate crops in the rotations could also benefit soil conservation. Secondly, Elzeard's co-development workshops bring together local practitioners from different backgrounds (teachers, farmers, advisers) and different practices (organic, conventional) in order to build a collective dynamic on diversified market gardening. The use of a common sharing tool, the Knowledge Greenhouse, could allow easier interactions and a common language between advisers and market gardeners around the theme of diversification.

To test some of these links, we conducted a sociological study in partnership with Ecole Supérieure d’Agricultures (ESA, 2019) about the profiles of market gardeners (smallholders to medium-sized, conventional to organic, peri-urban to rural…), their daily production needs, their difficulties and their actual use of digital technologies to help them. If we add the participants of the co-construction workshops in 8 French regions, we conducted over three hundreds interviews with the various players in vegetable production in France! We thus validated some of our hypotheses, and especially that well-organised farmers usually augment their productivity of 10 to 15% (and hence their incomes) but also gain 2 to 3 weeks of workload. 

For the moment we are testing our prototype with around thirty beta testers and we plan to increase this figure to reach 100 beta testers in September 2020.
The objective is to start selling The Market Farmer's Companion application this fall and to deploy it using our networks of prescribers (Chamber of regional and departmental agriculture, Organic networks, etc.) throughout France. 

Next year we aim to adapt our solution to other close countries such as Spain or Italy, the largest vegetable producers in Europe (80.000 farms), but also the United States and Canada (+75.000 farms) where the market already exists thanks to Tend and AgSquarred competitive solutions. We already have contacts and a beta tester in Quebec. 

The widening of our French target will allow us to reach +1000 active users at the end of 2021 and +10000 active users after 5 years. We will then deploy to farms situated in Eastern Europe, Africa and Asia.

Our Elzeard project is developing a platform of digital services and sharing of agroecological know-how for market gardening activities. Within the next year, we want to meet a double objective: to increase productivity and develop production that is more respectful of the environment. Thanks to Artificial Intelligence technologies, Elzeard provides a decision support service to offer optimal crop planning and assistance for the conduct of cultivation operations, based on local constraints and exploitable agroecological methods. This tool will directly impact the market gardeners using the application but also their workers and indirectly their families by bringing them more income and more time. 

In the medium term, Elzeard also wants to be at the service of the sector and to bring together the different actors of a territory through the sharing of practices, experiences and data within a collaborative network. Therefore, it will impact the external actors in direct relationship with the farm, including the networks of local producers, the technical advisors, the seeders and suppliers, the different distribution and transformation channels, the local authorities… We are actively developing that community of various actors by different partnerships.

In the next five years, our ambition is to bring digital innovation to this area and in support of productive agriculture in “agroecological” transition. We want to impact consumers by offering them healthier, gustative and sustainable vegetables with more transparency on their provenance. And on a more global scale, it will impact territories by relocating the vegetable production activities and developing short-route accessibility.

Within the next year, the main barrier to leverage is the limited adoption of The Companion application by the producers. This could be due to an unsustainable economic model forcing us to pivot. It could also be related to non-adoption of the tool by the farmers : not simple enough, cases too specific to provide reliable crop itinaries, lack of confidence in data sharing… To fight against these barriers we adopted a co-development process including the market gardeners from the begining in brainstorming, prototyping and testing activities. We also pay very attention to data management and protection. We are working to become RGPD and DataAgri compliants. 

During the next five years, another important barrier consist in financing the R&D of the projet. This means mobilizing important financing resources to hire human resources on the long-term to develop the ambitious features needed to complete the tools to offer maximum added-value to the producers. We also may have some issues related to the limitation of reasoning algorithms for planning and suggesting personalized crop itineraries. That is why we are developing partnerships with technical research teams to stay uptodate regarding AI methods and tools to implement in our reasoning engine.

Regarding the use-related barriers, we adopted a co-construction process of the tools with market gardening professionals and advisers to meet their needs as closely as possible. We know that to convince them of the utility to pay for the service we need to bring evidence about the gain generated by the application use (productivity, time, mental load...). We planned to measure impact opportunity with the help of technical agricultural experts next year. We also have to reassure the farmers on the modalities of data sharing by transparency: for whom, for doing what, in which circle…? This can be done by animating the community and setting up a transparent data sharing framework through the application of the GDPR+Data-Agri Code of Conduct (EU,2018). This legal barrier is overcome by active collaboration with standardization and data regulation bodies.

About the technology, the use of Semantic Web methods to interconnect structured (databases, spreadsheets) and/or unstructured (texts, images, videos) data already permits to reason and infer new knowledge. The other current choices are brought to proven technologies (REACT, JS, NoSQL, Graph databases…). 

Barriers related to economic and financial aspects can be overcome by workload adjustment (SaaS infrastructure, pace of recruitment, outsourcing recourse) and search for additional funding. We also rely on complementarity of partners sharing common interests and helping us on specific tasks. Nonetheless we set up a Strategic Committee to work closely with different representatives of the vegetable production chain to stay focused on the most strategic orientations for the project.

Here is our team:

- 4 full-time staff : 1 Admin & Business developer, 2 Full-stack developers (including an ex-market gardener!), 1 Senior AI/Data scientist (expert in Semantic Web)

- 1 part-time staff which is a UX designer

- 2 interships : one agronome specialized in vegetal production and one Knowledge engineer/Data scientist which will be recruited in September as a PhD student in collaboration with the University of Montpellier and LIRMM research lab. 

- 2 contractors that helps for front and mobile development 

- other : members of our Strategic Committee (5 producers, 4 researchers, 2 technical agricultural advisors)

To concretize the value proposition carried by the Elzeard solution, our team needs to master the technical fields: knowledge modelling and creation of agricultural data repositories; integration of recommendation and artificial intelligence services for decision support services; web design, agile development and Design Thinking for the most fluid user experience possible. It must also mobilize experience and skills: in agronomy for the integration of technical itineraries and business rules and work with our producers; in animation of network of professionals and network of partners; in marketing, digital communication and market development. 

The co-founders bring together all the key skills necessary for the project: IT, AI, knowledge management, data exploitation, Business Development, project management, animation of networks and commercial experience, human management… The Elzeard’s team members complete with resources in software development, agronomy, design and integration. 

And above all, we are developing our network of pilot farmers and partners (agricultural advisers, agricultural teachers, researchers) to integrate future users in a co-construction and co-development process. They feed our innovative development approach associated with participative governance to involve them in the definition of our strategy and guide our work. Our respective paths, our implication on the values of social and solidarity entrepreneurship, the desired impact around our solutions as our will to register in the development of a common good brings us to this conviction that beyond the raw skills, we have the assets necessary for the success of this ambitious project in this complex environment!

For 2 years, Elzeard has been engaged in a co-construction approach as close as possible to the expectations of actors in the vegetable production sector. We are developing our partners’ network around four axes: farmers and agricultural advisers, research and academics, local authorities and technical partners.

In particular, we work with market gardening experts from the French Agriculture Chambers or the organic farming associations. They help us identify and validate the reference data for cultural itineraries and find innovative economic models to relieve farmers. It is also necessary to study how digital tools can strengthen their support services to find together the best solutions to improve aid to producers in the field.

The mission of this partnership with the farmers and advisors is:

• to supply the Knowledge Greenhouse with knowledge and resources;

• to revisit practices in the light of current changes (ecological, digital and regulatory transitions) and to register new methods;

• share feedback, especially in rotation and cultural associations;

• propose usage scenarios and functional priorities;

• to criticize and validate the proposed interfaces;

• to test the prototypes developed and to carry out experiments.

We have also established partnerships with agricultural and IT research laboratories. We have been participating since June 2019 in the ANR D2KAB research project, coordinated by the Montpellier IT laboratory (LIRMM) and bringing together several INRAE labs, ACTA, WIMMICS team (INRIA) and API-Agro, specialized in modeling agronomic and agricultural data to build a portal to access the ontologies allowing interoperability and interconnection between data sets and digital services.

We have defined a hybrid model: the Knowledge Greenhouse portal is free and open while the Compagnon application takes the form of an all-in-one SaaS subscription. 

The Knowledge Greenhouse is free and open to allow sharing on a larger scale and to accommodate a diversity of resources. It foreshadows the creation of a common good administered by the community of vegetable production stakeholders. 

The Compagnon offers a wide range of services to farmers. Without any particular IT installation, automatically integrating update and maintenance services, it will generate savings in time and productivity of 10 to 15% (advisors interviewed). We offer a sharing of the generated value, with a subscription at a variable cost of 40 to 90$ per month depending on the size of the farm. We will also encourage word of mouth through a sponsorship offer between producers.

The planning module will be declined in an educational mode with an annual subscription around 1000$ for training centers. We want to provide their teachers with a support tool for making students understand the importance of planning and organization. This module will also ensure strong promotion among future market gardeners.  

Our market approach to our beneficiaries is also hybrid: physical+digital. We will participate in various agricultural events and fairs, combined with a permanent tour of prescribing support structures (Agriculture chambers, organic associations, cooperatives), co-jointly organizing workshops and presentations. Our digital action takes the form of a strong policy of content production, exploitation of professional networks and a virtual demonstration farm.

We apply to Solve Challenge for all the reasons mentioned in Become a Solver page! 

We want to address all market gardeners worldwide as that agroecological transition must be accomplished everywhere. We must adapt our technologies to be more resilient and aware of the urgency of the situation. Being part of the Solvers community will help us to share and communicate this message. It is important to be part of a larger community sharing the same desire and the same objectives to evolve towards a better future for all. It is a means of finding partnerships and synergies, enabling higher outcomes and therefore impact. We’d love to meet and get feedback from experts and mentors from the Solve community in order to test and improve our proposed solution. The great opportunity to participate in nine months of personalized support from Solve staff and Members of Solve’s cross-sector community is a unique chance to accelerate development and scope of our project. 

Alone we go faster, together we go further - African proverb

• Solution technology: to help us build the most adapted technical solution with latest research on Artificial Intelligence to offer powerful features to producers

• Product/service distribution: to find partners to distribute our application in northern america and world-wide as the challenge is international 

• Funding and revenue model: we need to raise funds to hire more staff (developers, community manager, sales, support, agronomes…)

• Board members or advisors: we are seeking for people ready to mentor and challenge us to find the best adequation between our revenue model and our social and environmental vision

• Marketing, media, and exposure: being exposed as a Solver can reach new beneficiaries, positively influence funders, develops partnerships, inspire new entrepreneurs for great projects and positive impact.

We obviously do not know all the Solve Members to target those who could help us the most. But we already identified the following organizations as very interesting to partner with: 

• MIT teams such as the MIT CSAIL or the MIT-IBM Watson AI Lab about new AI technologies that could be applied to our crop planification system but also the MIT Center for Collective Intelligence to work on the notion of digital inclusion, collective intelligence being one of the key to ease the appropriation of digital solutions by people. 

• Solve Members acting directly for food sustainability and agroecological transition such as the Foundation for Food and Agricultural Research

• Solve Members offering technical knowledge, support or services such as Attlassian Foundation Intal, Microsoft or Cisco to accelerate our work and scale our solution

• Solver Members that are consulting firms such as EY or incubators such as Antropia ESSEC to test our business model and help measuring our impact

• Foundations of course as we need to raise important funds within the next year to develop the full potential of our vision

We would like to qualify for the AI Humanity Prize as our solutions, the Market Farmer’s Companion and the Knowledge Greenhouse, rely on specific artificial intelligence algorithms and knowledge engineering for structuring, enriching and exploiting data to deliver a new kind of service for market gardeners.