Powered by Bank of Brazil and incubated in Silicon Valley, Agrobot was created to help small farmers make smarter production and business decisions, from planning the season to planting and selling. It does this by providing personalized advice on budget planning, weather, market opportunities and selling strategies.
Working from Sunnyvale, California, I lead the product design process from research, concept and prototyping to interface design and branding. I also contributed with a substantial portion of the front-end code in React Native.
Research
Concept & Strategy
Prototyping
Interface Design
Branding
Front-end Development
May – August 2017
Brazilian small farmers spend an average of 3 hours a day consuming agribusiness related content, especially about market, news and weather.
They live and breathe agriculture and are optimistic by nature. Still, in most cases, their decision-making process about budgeting, production and selling is based on intuition and guesswork.
Not having a clear picture of the production finances leads to higher risks, lower results and a harder time planning how to grow the business.
In our researches with small-scale producers, we found that the vast majority still use spreadsheets, notebooks and account statements to manage their budget, costs and estimate revenue. We also discovered that some don't even manage their finances.
This results in estimated production costs and revenue that are often inaccurate because they are based on guesswork rather than the current market dynamics. Not having a clear picture of the production finances leads to higher risks, lower results and a harder time planning how to grow the business.
Weather also plays a critical role in any agriculture activity, but it’s increasingly unpredictable, and the forecasts are often unreliable. While these are issues we cannot solve, we noticed that not planning for it increases harvest loss and lower product quality.
It’s also not always clear what is the best selling strategy. Sometimes it makes sense to secure buyers before planting, while in others holding on to the produce in expectation of higher market prices later in the season is a better move.
This made it clear to us how planning is key to maximizing profits and how critical it is to stay informed on prices, demand and processing bottlenecks.
Agrobot's proposition was to use Bank of Brazil's historical expertise in the sector and combine it with external services to provide farmers with personalized advice to help them make smarter decisions, reduce risk and optimize their results.
And our challenge was to create a thoughtful digital experience to support that goal.
Agrobot's target market is small-scale farmers within Brazil who need help managing production finances and achieving better results.
We used a combination of research methods, from online surveys to focus groups and contextual interviews, to better understand their behaviors and motivations, validate our hypothesis, and expose new needs and opportunities.
75% still use spreadsheets, notebooks and account statements to manage their budget and estimate revenue
Our surveys with 778 producers found that 75% still use spreadsheets, notebooks, and account statements to manage their budget and estimate revenue. We were also surprised to discover that 11% don’t even manage their finances.
Weather is also a critical aspect and affects every phase of the production cycle, from planning the season to selling and transporting. Not surprisingly, farmers check weather forecast an average of 6 times a day.
And while we (still) can’t solve weather unpredictability and the unreliability of forecasts, we learned that helping farmers plan for and approach weather challenges smartly could lead to more profitable seasons and better products.
In La Niña years, for example, dry weather in some regions reduces the incidence of fungal diseases. This results not only in better product quality and increased revenue but also in reduced cost with agrochemicals.
Here are some the conclusions we reached when reviewing our findings:
With a deeper understanding of our users and the main issues we needed to address, our team began ideating on potential solutions to each challenge in a production cycle:
Now it was time to start generating ideas for the features, starting with the onboarding process.
For Agrobot to provide any kind of assistance, it needed some key information about the farmer's enterprise, such as type of crop, plantation area, region, irrigation technology and workforce type.
We wanted the overall experience to feel like a friendly advisor, and the first moments were a great chance express that.
Onboarding flows are notable for their churn rate, so we knew that creating an objective and engaging initial experience was critical. We also knew that this was a unique opportunity to help users understand what the product was about.
We wanted the overall experience to feel like a friendly advisor, and the first moments were a great chance express that. With that in mind, we decided to explore a personal, guided conversational approach for the onboarding.
We were aware that a conversational onboarding could feel confusing, so we wanted to test it with real users as soon as possible. We experimented with creating prototypes using Botmock or Botsociety, but the interactions didn't feel accurate enough for our needs. Luckily we were able to quickly come up with an actual, coded prototype and test it.
From a usability standpoint, our onboarding prototype proved successful. We were primarily interested in testing if people would understand and answer the questions quickly and have a clear understanding of the product's purpose.
Now that we had found a simple way to onboard users and get the key information Agrobot needs to start assisting, I shifted my focus to exploring how the product would advise farmers in the different phases of their production cycle.
Weather and market information are checked very frequently by farmers, so it made sense to display them in the most direct way possible, right on the home screen.
But features accompanied by some specialized analysis such as price tendencies for future contracts or detailed alerts for weather and diseases anomalies felt more comfortable in a conversational interface.
With that in mind, I started studying ways to display frequently requested themes such as market and weather monitoring in a dashboard and conversational functionality in a chat interface.
The following designs illustrate how we ended up providing an overview of the crops in a dashboard, along with widgets for frequently requested topics such as market and weather monitoring. Advisor functionality that worked best as conversations were centralized in a chat interface.
Agrobot is informal and friendly but also powerful and objective. The design language had to express that while also helping users feel productive and confident to make decisions.
Space was used plentifully to express comfort, organization and simplicity, while interface elements were carefully organized to make sure the purpose of each screen was immediately clear.
A good portion of the interactions with Agrobot happens through conversational interfaces, so a great voice and tone strategy was especially important.
Typography choices prioritized clarity and friendliness — a warm typeface combined with texts appropriately varied in weight and color helped create a strong hierarchy and familiarity.
Interface elements are usually flat and simple to make the interfaces clear and objective, but light shadows were used to express elevation and indicate elements that support touch or gestures.
A good portion of the interactions with Agrobot happens through conversational interfaces, so our voice and tone strategy was especially important.
Agrobot talks in an objective way, but its voice is always familiar and kind. And while it may sometimes surprise you with unexpected wittiness, it knows that the content and clarity of its message are more important than showing off its personality.
To help us make unified and consistent design choices, I created a Design Persona. It described Agrobot's personality and how it manifested in the product through visual design, copy, and interactions. You can check it at the link below:
A variety of modern development technologies were used in this project, including React Native, MongoDB, Dialogflow and Node.js. I was responsible for a substantial portion of the front-end code in React Native collaborated with two other full-stack developers.
Working as a designer and developer was efficient and rewarding since I had greater control of the final product, but it was a challenge to switch between developer and designer modes.
Agrobot was a very rewarding project, mainly due to its inspiring mission and intriguing problem space. It allowed me to work from Silicon Valley and think through business problems and user contexts that were entirely unfamiliar to me.
With a small team and a condensed timeline, I strived to organize my time and prioritize tasks to ensure every design discipline received appropriate attention; from research, concept and strategy to prototyping, interface design, testing and front-end development.
In 2018, Agrobot won the Cantarino Brasileiro Award in the Artificial Intelligence category. Recognized as the main award in the financial sector, the prize highlights the best contributions to the Brazilian financial industry.
Although we didn't have the chance create a comprehensible Design System specification, we managed to build reusable UI components and patterns that allowed for a cohesive design language throughout the product, while also maintaining a consistent voice and tone.
By the time I left the project, not all of the designed features were released, but Agrobot had over 3000 users in 23 different states and the team was in the process of tracking key metrics, including:
