Chatbots made in Africa: collective action and lessons learned

How to get started with chatbots workshop at the 2020 NetHope Global Summit.

Elizabeth Njoroge, International ICT Business Partner, Christian Aid, Ahmed Mihaimeed, Regional ICT Director, SOS Children’s Villages, Mazin Elamin ICT4D Coordinator, SOS Children's Villages, Doreen Koech, Information Security Lead, VSO

While AI has now been around for a long time and is forecasted to add more than $15 trillion to the total global economy by 2030, this remains a mirage in Africa. Only a few countries are making use of AI effectively in Africa. The Global South risks missing out on the Fourth Industrial Revolution if it does not take advantage of the opportunities that AI presents.

One of the biggest challenges to effectively utilising AI in Africa is the skills required to design, develop, and maintain AI-enabled systems.

One of the AI applications with a broad applicability in the social impact sector and with the lowest barrier to entry thanks to no-code and low-code tools is chatbots, or software applications that use conversational AI (Natural Language Processing) to understand what a human wants and provide meaningful answers. Chatbots are quickly becoming a popular method for organizations to interact with the communities they support and their staff.  They can help nonprofits reach more people with the services and information they need, can respond to repetitive information requests, and more.

Never has the world needed partnerships and collaboration to transform people’s lives more. As the NetHope Africa Chapter, we have been discussing chatbots for some time, and were keen to come together collectively and use this technology to benefit our work and those we are working with.

We asked NetHope to support the Africa Chapter to explore the application of chatbots in the continent. In collaboration with Microsoft, NetHope put together a package that included project-based training, technical and domain expertise, seed funding, and facilitation of the projects from start to finish for two chatbot concepts.

Made in Africa: our two chatbots

Knowledge Sharing Chatbot (Mama Africa)

• Lead NGO: Multiple NGOs from the Africa Chapter, led by SOS Children’s Villages, with support from VSO and Christian Aid
• Problem: Access to technical training and IT-related knowledge (product, partner/vendors). No way to share the vast knowledge within our Chapter with all Chapter members
• Target audience: IT/ICT staff at NetHope Member organizations
•  Solution: Connect community with knowledge:
  • Provide a portfolio of training content, information, and NGO best practices across different technical areas (C&I, Cloud Services, Governance, etc)
  • Connect the users to the local expert network
  • Enable experts to contribute content to the bot
• Data: Initial content provided by core team. Crowdsourcing of content through the Chapter experts group
• Channels: Website, Microsoft Teams

Tech Support Chatbot

• Lead NGO: Christian Aid UK
• Problem: Access to technical support for geographically distributed staff, with reduced support workforce. A need to streamline and automate support processes
• Target audience: All staff at Christian Aid (both technical and non-technical, HQ and field)
• Solution: Encourage and equip staff to resolve technical issues:
  • Provide access to technical support knowledgebase via a self-service, chat-based functionality
  • Provide access to human-based support Jira ticket submission through Power Automate
  • Provide ability to request equipment
• Data: Knowledgebase includes information from Confluence, Online content and crowdsourced content.
• Available via Microsoft Teams

Africa chatbots: lessons learned

1. Management buy-in and expert support is essential to the success of the project.

• The key to chatbot success is buy-in from the top which leads to commitment at all levels of an organization. For example, through commitment from their CTO, Christian Aid provided an environment where it was safe to test, fail and learn.
• Internal resources/team: Building a robust chatbot implementation is a team effort. Some of the responsibilities include content gathering and curation, script writing/editing, bot development in PVA, testing with end-users, and more.
• Partner support: The ongoing support and guidance from NetHope and Microsoft enabled us to not only develop the bots but also build technical and domain capacity at the same time.

2. Content is everything.
Chatbots are as useful as the quality and accuracy of the information they provide, so it’s critical to curate a high-quality content library and develop a sustainable content refresh model. For example, for Mama Africa Chatbot, content will be crowdsourced from the NetHope Africa Chapter community. We have asked all member organizations in the Africa Chapter to contribute relevant information with the goal of curating a diverse knowledgebase that is representative of various experiences across the continent.

3. No-code or low-code tools make it possible for anyone to get started with AI.
No-code and low-code AI tools are key to democratizing the access to AI capabilities, enabling individuals and teams to get started without having any specialized expertise. With Microsoft’s Power Virtual Agents, a no-code tool for creating chatbots, you can build a simple chatbot within minutes. More complex use cases require more work. For example, the Tech Support chatbot required integration with Jira as the ticketing system and Confluence as a knowledge management system. Michael from Microsoft was quick to provide technical guidance and bring in engineers for additional support, so we were able to implement Jira connector without a single line of code and build the know-how that we can pass onto other nonprofits.

4. Design responsibly from the start.
With support from Leila at NetHope, we laid the foundation for responsible design early on by taking the time to assess the need and purpose for the bots (vs using existing solutions), develop a good understanding of target users, articulate the unique value proposition, and map out how the users will engage and benefit from the bots in their contexts (ie, user journeys) and what might be barriers to engagement (eg, connectivity, devices, digital literacy, non-inclusive language). Throughout the development process, we engaged in the discussions about ethical considerations. This consequence scanning enabled us to anticipate and address potential ethical issues. For example:

• To ensure that the chatbots treat all people fairly, we used NetHope’s AI Ethics toolkit to consider questions of equity, representativeness and bias, which helped us to optimize: (1) bot persona and script; (2) content and features; and (3) implementation for a more fair outcome for all. For example, Tech Support Chabot was initially named Christian. While this is a person’s name, we have staff from all walks of faith and we realized that this name could leave out non-Christians, so we changed it.
• To ensure privacy concerns are addressed in Mama Africa chatbot, we decided not to gather certain information and to include proactive communication about data collection (eg, user email) to build trust.
• To provide transparency for Mama Africa Chatbot, we created an onboarding video that proactively explains the purpose of the bot, how it works, the user role (eg, learner, contributor) and that communicates that it's a machine, not a human.
• AI systems should empower and include everyone, which in the case of Tech Support Chatbot (serving Christian Aid staff across different contexts) meant considering how to design the bot with inclusiveness in mind, eg for people with low(er) digital literacy, limited access to connectivity, or providing support with connectivity/devices/power. And we tested the bot with those end-users.

Finally, we’re excited to share our bots as templates that can be adapted and reused by anyone. We hope this will encourage more nonprofits to consider adapting and building on what works and accelerate our collective impact.

Join us for a NetHope Solutions Center webinar on June 15^th to see the chatbots, get access to bot templates, and hear more about the lessons learned. Sign up here to receive the webinar invitation.