Our 1st chatbot design sprint with… ummm… farmers in Kenya

Our experiences (and some methods) for testing chatbot prototypes with farmers in Kenya, which worked better than you might expect!

Over the past few months I’ve been following articles and insights from Chatbots magazine and playing around with chatbots on botlist with interest. Just like anyone else I’m excited to get onboard the chatbot revolution but my interest is — perhaps — a little different from the norm.

In a nutshell, I’m interested in whether we can make chatbots work for farmers in Kenya. If that sounds a bit silly then read why we think making chatbots for farmers in Kenya isn’t as stupid as it sounds. Hopefully it will convince you.

There aren’t many chatbots in the farming/agriculture space, let alone in Kenya, so we’re learning a lot as we go. In this article I want to share the story of our first design sprint prototyping chatbots for Kenyan farmers.

Our approach (and a new discovered love for Post-it notes)

First things first. When I say a ‘design sprint’ I mean the process outlined (amazingly well) in Google Venture’s “Sprint” book. We also dipped into IDEO’s “Human Centered Design Toolkit” (a great cook book for interview techniques, and ways to prototype solutions) to give us inspiration.

Outline of the method (copied from the “Sprint” book)

We’d already reviewed a range of chatbots drawing some guiding lessons from them and a mini directory of examples to “remix and improve” on. By having a blend of expertise in the room from software development, farming, Swahili, data wrangling and more, we were able to map out the journey of a Kenyan farmer in incredible detail and brainstorm chatbot solutions. There was also huge value in having the collective thinking of the team literally stuck to the walls (hence the new discovered love of post-its)

Our prototype room

The idea of the sprint method is pretty simple: Come up with a product (in our case a chatbot), make a prototype that your target customer interviewees believe is the real-deal (in our case Kenyan farmers who have Facebook), and do it all in 5 days.

Our Farmer Chatbot testing lab

So how did it go? It turns out prototyping chatbots has a hidden advantage. As IDEO point out this might be because conversation is itself a design tool.

The last day of the sprint is for running customer interviews with a prototype product. Most of the team is in the sprint room observing customer interviews passively. However, if you’re testing a chatbot there’s a natural opportunity for this team to get more engaged.

Diagram of the interview setup (copied from the “Sprint” book)

For example, we played around with subtle personalisation of messages, changing the order of messages, and so on. While it’s important to stick to a rough script, these kind of subtle tweaks to conversation can be great learning experiences. They’re also very natural since you’re not pretending to be an app, you’re pretending to be a chatbot (which in some sense is pretending to be you). While there are great tools out there like Chatfuel and Motion.ai, our experience is that going manual is better for prototyping. (Actually, I think this applies just as well to chatbots targeting Kenyan farmers to those targeting New York fitness fanatics.)

To get all this working we used the following tools for our “research lab”:

  1. Two Macbooks linked via Skype — so the sprint room team could watch the interview live
  2. Postman — so a member of the sprint room team could send chatbot messages manually (via POST request to emulate a bot on FB messenger)
  3. Lookback — to record the user screen and touches to get a solid record of the customer experience for us to analyse
Our setup for the chatbot prototype

Some Insights

So what did we learn from all of this?

The first basic (but important) validation is that chatbots are totally viable solution for farmers. You might not believe it but Facebook has been a game-changer for droves of Kenyan farmers. As one farmer put it “Facebook and IT has opened up a very big door”. None of the interviewees had ever heard of a chatbot but they were all very proficient operating one (since they spend as much time as any of us on WhatsApp or Facebook Messenger). Every user seemed genuinely excited about using the chatbot and getting new services over a channel they were already very familiar with.

“Facebook and IT has opened up a very big door”
Feedback on our prototype chatbot

The second big finding is that interviewees were excited about multiple features. They also clearly benchmarked the value propositions against digital behaviour they were either aware of or engaged in. As one farmer mentioned, “Facebook used to be about communication, but now it’s about marketing”. Getting connected to the market is the biggest challenge, but timely agronomic advice in emergencies was discussed, in addition to better market information to plan crop planting cycles. There are lots of variables in farming and a dearth of good information. Digital channels are doing some good work in meeting the demand for better information, but a lot more could be done.

“Facebook used to be about communication, but now it’s about marketing”

Finally, one of the most of interesting features is the role of groups. Farmers are part of large digital group discussions about farming on WhatsApp, Facebook, Telegram and more. This is where most of the value is gained but also where most of the frustrations are felt by users. Augmenting the conversations and behaviour in these groups looks key (though incidentally, this is where chatbot prototypes become a little harder… trying to prototype person-to-person conversation is a bit easier than person-to-group conversation… one to mull over)

Closing Thoughts

Hopefully all this was vaguely interesting because “chatbots for Kenyan farmers” isn’t a topic you read much about. This aside, I think there are some bigger takeaways for those who aren’t tracking #chatbots4farmers religiously (those who are definitely please get in touch with me)… Here are two closing thoughts for the rest:

Should chatbot makers pay more attention to users in developing economies? Everyone who gets excited about chatbots also point to the phenomenal growth in use of messaging platforms. But where is this growth coming from? We all know that it’s largely developing market economies. And, let’s face it, there isn’t much known about what users in places like Kenya want from a chatbot. There’s a lot we could learn from the rapidly emerging East African mobile messaging generation. We should all start paying more attention to this set of people not just because of it’s size, but also because the space of problems to solve for it is much bigger: from agriculture to education to financial services. Indeed, perhaps chatbots for these people will have an easier ride transitioning from novelty to utility in light of greater needs and a dearth of alternatives.

Can service providers in developing economies catch on quicker to chatbot opportunities? It’s totally feasible to take the best practices and technologies behind chatbots and make solutions for a target market in East Africa. There are upward of 7.8 million smartphones registered on Safaricom’s network in Kenya right now and 5.8 million Facebook MAUs growing 18.6% from last year (that’s not counting any extras for other mobile network operators or other mobile messaging platforms). Moreover, as I hope this article demonstrates, we can take methods like ‘Sprint’ and apply them perfectly well in Kenya and other markets. Those working in the Kenyan market will also realise that it’s more of a group conversational culture than a solitary app culture, and that looks good for chatbots. Also there’s a great chance to start at the same time as those developing chatbot services in the U.S and Europe, enabling early movers to get a competitive edge. If you’re in Kenya and were thinking of making an app, maybe you should be considering a chatbot…

If you’re interested to follow what we’re getting up to building chatbots for farmers you can follow us at farm.ink on medium. I’ll aim to post more about our progress and insights as we go, and would love to hear from anyone thinking about chatbot solutions in East Africa in particular. Also if you have comments on the thoughts in this article please do share them, as it would be great to discuss. Thanks for reading!

Enjoyed the article? Click the ❤ below to recommend it to other interested readers!
One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.