Before AI became a thing that everyone talks about and wants in their products, there were people already innovating in this space, case in point, Irving Amukasa.
Hi Irving, what’s your background, and what has been your career?
Hi, I have had a pretty unconventional path to entrepreneurship. Straight out of high school, I taught myself how to build Android apps. I then applied to my first-ever incubator with my third-ever app thinking it was a contest for techies. A program called the Safaricom Appwiz Challenge.
Even though I did not get funded at the end of that program I learned a lot of entrepreneurial skills. Business modeling, financial projections, hustle, I didn’t even post that app on my own Play Store link. Fast forward to 2016, I am disillusioned with school as I go into my first year-long holiday and there is a call to action asking for techies to try and solve for sexual health.
I applied with a background of playing around with rudimentary chatbots and we got funding from UNFPA Kenya. I haven’t looked back since.
What are you working on currently and how did the idea come up?
I run SophieBot.
SophieBot is an AI chatbot to answer your questions on sexual health. Yes, I just Said Sexual health. On this side of the world, it’s awkward to have these conversations openly and honestly without taboo or sensationalization, we hope a non-judgemental anonymous chatbot can drive open and honest conversations on the topic.
The idea was from the intersection of my playing around with rule-based chatbots in 2015 and a call to action by UNFPA in 2016 for techies and innovators to try and solve for sexual health.
How did you build it?
Earlier on our tech was rule-based where you had to predefine an expected question and then provide an answer for it. Those rules were written in an old Markup language called AIML (Artificial Intelligence Markup Language).
We have now evolved our technology alongside advances in AI and we now use our own custom transformer AI model to answer new questions after being trained on old questions. Basically the T in chatGPT. We have our own.
How did you get your first customers/users?
I am a huge believer in organic growth. Then using those impact figures to get free press. My first 250 users were from a community of Android developers I helped run Android254 and the guys I was on campus with. I even remember the JKUAT Tech expo let us co-opt an exhibition stand from a sponsor who didn’t show up to the expo.
These efforts were helped along by the fact we give out and still give out a lot of branded material. Wristbands, Stickers, the works. We then capitalized on that growth to get free press. We are still up on the JKUAT site I think. Most of the Dailies have featured us in a way. We even got a copy of our local airline carrier’s inflight magazine, Msafiri.
What are some challenges that you have faced and how did you overcome them?
Our technology wasn’t as great at answering questions for a long time. We had a default answer for when someone asked a question not among our rules, “Hmm let me think about it”. The reviews from users were not nice at all. Most of our existence has been looking to solve that problem. AI that learns from previously expected questions to answer new unexpected questions.I am glad that we now have that deployed.
Financially/user acquisition wise how is the product doing?
We Just got asked 3000 questions in the last 3 weeks in all the channels we are on. We are growing both our user count and question count evenly at 10% more week on week. Most of our attention is coming through our new SMS and WhatsApp channels, something we are excited about. We also obsessively monitor the ratio of users to questions to see on average how long a conversation with a user lasts and it’s at 1:5 now and steadily growing.
What are some of the mistakes you made & some disadvantages of working on the product?
It is now entrepreneurial dogma about how large or small your founding team should be in order to be successful. I fell into the trap of recruiting a large team in 2016 and not constantly reiterating my vision to them. This led to a point where we got differing visions of where the company should go.
A lazy menu-driven sexual health content USSD menu or we keep at trying to solve answering questions in natural language. I prevailed on the vision but had to buy out my other shareholding co-founders.I could have spared us a lot of grief.
What are some of your future plans?
We just got access to more resources so we are training a larger SophieBot model that will be able to retain context across different questions, something AI folks call Token Length.
We also want to offer the same tech that answers questions for our users as a customer support tool for our B2B partners. We can now train our AI model on past conversations to answer new unexpected ones.
If you had the chance to do things differently, what would you do?
I would not change a single thing, all the grief and heartache too.
Any recommendations in terms of who to follow, YouTube videos to watch, a useful website or books to read?
For Folks trying to get into AI, Andrew Ng’s Stanford Machine Learning Lectures are a strong foundation. Andrej Kaparthy’s blog is very resourceful for folks who want to get into our type of AI, Natural language processing.
For Entrepreneurs, the Lean Startup book is a goldmine to build in the least wasteful way with the least resources testing all your assumptions as you go.
I alongside 46 other entrepreneurs also feature in a book about how to build in Africa, Model47. A copy is out on Amazon.
Lastly, Wisecrack on YouTube is a good resource on how to use philosophy to look critically at pop culture, media, and current events.
Where can people go to know more about you and your work?
My personal account is @iamukasa on nearly all platforms. Aside from shenanigans, I religiously post progress and updates there.
To find out all the channels SophieBot is deployed on checkout sophiebot.ai
Bonus question, What does success look like to you as an individual and at what point would you say you have achieved said success?
We want to be the largest resource on sexual health. I also want to build out our customer business to a size where you can exit the business, sell the company, or list shares publicly. Get a little more liquidity to fund other founders who look like me, and have the same background as me in this ecosystem.
Thank you for reading this far. If you have enjoyed today’s edition do share with a friend to tell their cousin. Subscribe to keep supporting the work.
