There is a lot of enthusiasm about the use of digital technologies to solve market failures in agriculture. A forthcoming report by the World Bank finds that by reducing transaction costs throughout the agro-food value chain and enabling increasingly customized advice on-farm, they can support more efficient, equitable and environmentally sustainable food systems. Personal experience, however, suggests that we should manage our expectations when it comes to seeing tech as the solution to all our problems. If you’ve ever spent a whole day trying to connect a printer to a computer or lost hours of work because your computer has crashed, then you’ll understand what we’re saying. Because. Just like some printers, not all technologies out there are mature enough to fix our food system.
So, what does this say about a project aimed at using digital technologies to boost the production of cashew nuts in Benin, West Africa? Spoiler alert: you need to be creative and ruthlessly persistent.
. With the support of TechnoServe, a Washington-based NGO, the country aims to digitize its cashew sector by deploying big-tech for small-holder agriculture in order to organize farmers and increase quality.
The first step of this initiative was the launch of a digital library of best agricultural practices for cashew cultivation. The idea seemed simple: the library would be accessible to extension workers across the country through an automatic mobile chat function, or what we commonly know as a chatbot.
But how do you connect an extension worker in Benin’s Borgou region to a Google-enabled language-processing engine hosted in the US?
Anyone who has recently visited an online store knows that chatbots are everywhere nowadays. These artificial communicators have evolved from an annoying pop up on a webpage to increasingly useful tools to connect users to information. So why not bring this technology to extension workers in Benin?
But where do you start and which providers should you partner with? There are many tech providers out there, and they all claim that everything is technically feasible. But it is only when you build your first minimal viable product (MVP) that your most basic assumptions get challenged. This is when you start dipping your toes in the icy waters of “implementation”.
Questions will start unfolding: Is there internet connectivity? That is an obvious one, but the devil is in the details. For instance, how do you explain complex agricultural practices in the 140-character SMS messages? In many ways, extension workers are not much different to online shoppers. They are picky about usability and you will soon have to understand that. This sets you up for user testing.
The good news is: Testing is easier than you may think and sometimes you just have to get creative, like pretending to be a chatbot (a so-called pretendotype) to simulate and learn how users would interact with the technology. Recording, measuring and learning from these experiences allows you to understand user needs--as well as to learn that some usersflirt with an agriculture chatbot while others will inquire about job opportunities.
Lots and lots of iteration
In Benin, one main challenge was to provide chat services across a variety of communication platforms. It took around ten separate MVPs to identify the right service providers and user interfaces and to launch the first public version of the BeninCajù Chatbot on SMS, WhatsApp and Facebook. Surprisingly, the oldest and most reliable channel –SMS—proved to be the most difficult to deploy. Providing automated two-way SMS communication at a reasonable cost seemed impossible. The most prominent actors in the chatbot space, such as Twilio, have not yet entered many African markets and the available alternatives in Benin were way too expensive.
Once again, the solution required a bit of creativity and rewiring existing solutions. We ended up hosting a SIM card virtually in another country and forwarding messages half-way across the globe to get the right responses to extension workers on the ground. This allowed the technology to build on Benin’s existing telecommunication infrastructure, and also allowed bringing in services that local telecoms were unable to offer.
First, an obvious step: Governments need to build tech-supportive environments and allow innovators to inform policy. They should ensure that the country has a competitive telecommunication sector and market access for a range of services provided from abroad.
Second, a logical step: Governments need to invest in institutional innovation to create networks of stakeholders. This supports public-private partnership and links use cases to technology providers, often revealing missed market opportunities and attracting technology players to their countries.
And finally, an often painful step: Governments and donors who invest in innovation need to be prepared for failure along the way. Failure is part of successful tech projects, provided that failures lead to learning and improved outcomes. Governments and donors can help defining desired outcomes – leaving creative methods to entrepreneurs – and by creating channels for sharing knowledge that support building experience in scaling technology for the agro-food tech sector.
Similar to agricultural value chains, the digital value chain for Africa’s tech revolution also needs fixing. Policy action can attract tech companies to new business opportunities, ultimately allowing tech providers to prove their capabilities on the ground and spill-over knowledge to local players.
We hope to crowd-in some of the world’s best minds to participate in a global conversation on food and technology through the “What’s cooking? Rethinking farm and food policy in the digital age” blog series. We invite people with diverse backgrounds and perspectives to join us and share their comments.