Artificial Intelligence to Identify Pests and Generate Agricultural Input Recommendations

AI solution to optimize diagnoses and recommendations in agribusiness

Role

Product Designer

Industry

Agribusiness, Artificial Intelligence

Duration

2 months (August and September 2024)

Step 1: Define the target audience and analyze challenges faced

In the first stage, the target audience was defined: agricultural consultants who use WhatsApp as their main working tool, but prefer paper records and face difficulties with complex systems.

We encountered some challenges that the target audience faced, such as:

Inefficiency in formalizing commercial proposals.
Difficulty in centralizing and sharing documentation.
Imprecise diagnosis and recommendations, impacting productivity.

Step 2: Project objectives

The main objectives of the project were then defined. The first was to optimize repetitive interactions, bringing more efficiency and assertiveness to the consultants' daily activities.

The second aimed to provide reliable diagnoses and recommendations for the use of agricultural inputs, helping to identify potential problems in plantations at an early stage.

And the third objective was to scale up learning about new inputs and treatment methods, democratizing access to knowledge through an intuitive and effective solution.

Step 3: Research and Insights

The third stage was dedicated to research. Initially, we used brainstorming sessions to explore ideas and broaden our understanding of the problem. We then benchmarked competitors, gaining valuable insights into the market and its main products.

To align the team, we applied the Problem Statement framework, ensuring a clear and shared understanding of what we were solving and for whom the solution would be developed.

Step 1: Define the target audience and analyze challenges faced

In the first stage, the target audience was defined: agricultural consultants who use WhatsApp as their main working tool, but prefer paper records and face difficulties with complex systems.

We encountered some challenges that the target audience faced, such as:

Inefficiency in formalizing commercial proposals.
Difficulty in centralizing and sharing documentation.
Imprecise diagnosis and recommendations, impacting productivity.

Step 2: Project objectives

The main objectives of the project were then defined. The first was to optimize repetitive interactions, bringing more efficiency and assertiveness to the consultants' daily activities.

The second aimed to provide reliable diagnoses and recommendations for the use of agricultural inputs, helping to identify potential problems in plantations at an early stage.

And the third objective was to scale up learning about new inputs and treatment methods, democratizing access to knowledge through an intuitive and effective solution.

Step 3: Research and Insights

To align the team, we applied the Problem Statement framework, ensuring a clear and shared understanding of what we were solving and for whom the solution would be developed.

Stage 4: Opportunities

We then explored the opportunities for applying AI in the context of the company.

Among them were the design of a recommendation engine to improve the personalization of suggestions and the development of a customer service tool to create scheduling reports.

Step 5: Creating the MVP

The minimum viable solution (MVP) developed at this stage included the creation of user flows and interactive prototyping. This phase aimed to validate the value proposition and usability of the solution with agricultural consultants before moving on to development.

Stage 6: Proof of Concept

We then carried out a proof of concept using the GPT What is This plugin, an AI tool trained to identify objects in images.

We tested the system with images from the soybean disease identification manual. Although the AI was correct in identifying the plant, it made mistakes in the diagnosis, indicating another disease.

However, the tool showed learning during the interaction, requesting more information to improve its predictions and adjusting its management suggestions based on the new data provided.

Step 7: Analysis of Future Opportunities

Finally, we identified future opportunities to improve the solution, including:

Improving AI Accuracy: Continuous training with local data.
More Precise Recommendations: Inclusion of climate data, soil analysis and crop history.
Availability to the End Customer: Evolution towards a self-service model.

Reflections

This project highlighted the importance of deeply understanding the target audience, adapting solutions to the context of use to ensure their effectiveness. The combination of research, continuous validation and multidisciplinary collaboration was essential to developing a tool that balanced innovation with simplicity, while respecting the needs of agricultural consultants.

The proof of concept highlighted the potential of AI as a resource that learns from users, but also revealed the importance of continuous training to increase the accuracy and reliability of recommendations.