Made in Africa AI in MERL Landscape Study: Reflections from Preliminary Findings 

Since April 2025, I’ve been leading a landscape study at The MERL Tech Initiative to explore how artificial intelligence (AI) is shaping Monitoring, Evaluation, Research, and Learning (MERL) across Africa. The goal of this study is to deepen understanding of the needs, gaps, and priorities of the AI+Africa Working Group at the NLP-CoP and related communities of practice to inform the broader AI in MERL landscaping work. Through this study we are also understanding not just the potential of AI in MERL, but also what it means for AI to be “Made in Africa”.  To do this, I conducted Key Informant Interviews (KIIs) with 20 diverse participants—practitioners, NLP developers, and researchers working in diverse contexts across the continent.

We wanted to bring the findings to our AI+Africa Working Group—not just to share learnings, but to invite collective reflection and feedback. This collaborative discussion is a key part of my research methodology to create shared ownership of the final product- learning rooted in, and led by the community. 

On June 24, 2025, I presented the preliminary findings from our ongoing Made in Africa AI in MERL study to the AI+Africa Working Group. My aim was to present what I’d learned so far, validate these early insights, and spark a conversation that would shape the research direction moving forward.

Can AI accelerate MERL efforts in Africa, and does it do so when it is developed locally? What tangible/practical use case evidence exists for the effectiveness of AI in MERL? What are some of the challenges that would benefit from the adoption of a Made in Africa AI in MERL Approach? During this meeting we attempted to answer some of these questions during an interactive feedback session with the  AI+Africa Working Group.

Preliminary Findings

Our first round of qualitative research findings is based on 20 stakeholder interviews, highlighting eight key themes:

Capacity and Readiness

Limited technical knowledge, a lack of context-specific training, and inadequate digital infrastructure continue to shape the trajectory of AI integration in African evaluation. It’s clear to me that foundational AI literacy shouldn’t be an afterthought—it needs to be woven into evaluation education and strategies from the start. This is essential if we want our systems to be truly ready for AI rather than treating it as an optional add-on, given its direct impact on readiness.

Defining “Made in Africa” AI in MERL

The definition of “Made in Africa AI,” as reflected in the findings, encompasses data that authentically represents local realities, languages, and knowledge systems. A Made in Africa AI approach in MERL is grounded in community-driven, participatory methods that prioritize indigenous definitions of success, rather than relying on Western-centric, colonial metrics.

Building on African Evaluation Frameworks

It emerged that rooting AI initiatives within existing African evaluation frameworks, and emphasizing culturally and ethically aligned practices, is key to ensuring that AI integration mirrors lived realities. Movements like #FeesMustFall show us the power of homegrown, community-led change—and offer valuable lessons as Africa shapes its own technological future. 

Transformative Potential of AI in MERL: A Unique Opportunity for Africa

Optimism was evident regarding AI’s potential to significantly enhance MERL effectiveness in Africa. Examples included the development of rapid assessment tools tailored for emergencies and the use of chatbots to boost youth participation. This transformative potential is also linked to lowering entry barriers for key groups, such as African youth, as a means to fully realize this transformation on the continent.

Climate Implications of AI in Africa

Concerns about the environmental and social impacts of AI in Africa featured prominently in the findings. Specific reference was made to resource extraction and its link to conflict and turmoil in countries like the DRC. If we want AI innovation to benefit Africa, we must embed climate mitigation strategies from the outset—otherwise, we risk repeating old patterns of exploitation.

Appropriateness of AI Tools for African Data and Diversity

While practitioners widely use AI tools for evidence gathering and analysis, it emerged that many existing tools are ill-suited to African contexts. They often fail to adequately handle African languages and rarely support the qualitative richness or local data structures that MERL practitioners require.

Growing Use of AI: Potential Risks and Harms

Concerns surfaced around the dangers of AI and technodeterminism. Issues such as data exploitation, privacy breaches, loss of meaningful human engagement, and algorithmic bias were cited as growing trade-offs with the increasing use of AI in MERL. As we lean more on AI, we need strong governance frameworks to protect our interests and ensure technology serves—not harms—our communities.

Strategies for Collaboration and Movement Building

There was consensus on the need for collaborative, safe, and inclusive spaces where practitioners at all levels can learn, experiment, and share experiences. The recommendation is clear: AI should be integrated into the mainstream of evaluation practice, and not relegated to a niche subject for a select few.

What we learned together: Insights from Working Group Members

After sharing early insights, we gathered feedback from those who had joined the call. Their feedback on local AI policies, especially within public institutions and NGOs, highlighted the importance of including a greater focus on gender gaps and limited capacity among youth and women in the final report. In the next phase of the study, I will delve deeper into these governance gaps to identify what is needed to advance AI in MERL across Africa, particularly within public institutions, which play a central role in setting AI-related policy and frameworks.

Questions also arose regarding practical use cases and whether Africa should focus on developing new AI models or improving existing, smaller-scale ones. There was a strong argument for MERL practitioners to first become advanced users of existing tools—fine-tuning and adapting them locally—rather than investing heavily in entirely new systems. Given the high costs associated with building AI (often in the millions of dollars), this approach was presented as a pragmatic entry point for the continent.

In defining a Made in Africa AI in MERL approach, it is critical to frame the research within a human rights framework, i.e. the integration of economic, gender, and epistemic justice considerations alongside climate concerns, particularly in addressing inequalities perpetuated by AI supply chains and biased data systems. AI and climate discourse often creates a paradox, as it competes with systemic issues such as corruption, extractivism, and greed.

Critically examining language and terminology used in evaluation plays a huge role in how a Made in Africa AI in MERL approach will be framed. This is particularly relevant to terms like  “vulnerable communities,” and how we need to shift from such practice and prioritize language that respects autonomy and creates agency.

The group addressed the growing hype around AI adoption, arguing that MERL practitioners play a key role in challenging this narrative. AI is often seen as a “hammer” that can solve every problem—an approach that can be problematic in evaluation. African MERL practitioners need to scale up their engagement, influence the design of datasets and prompt matrices, and help shape the models used in practice.

Looking ahead, deeper reflection on the differences between generative AI (currently mainstream) and agentic AI (on the horizon) is pivotal, and what these developments mean for MERL in Africa. AI “sees” the world through tokens, which complicates public understanding of what AI can reliably do, especially in evaluating sensitive programs.

My Reflections

I see the opportunity that AI presents for MERL and acknowledge the rapid pace at which emerging technologies are evolving, compelling Africa to participate and define its own approach. There is great value in understanding existing AI tools and translating that knowledge into frameworks that respect and uplift our indigenous knowledge systems.

However, the current AI hype can hinder how Africa shapes, adopts, and develops AI in a way that truly reflects our collective understanding of how these technologies should serve the continent. For me, this moment presents two key opportunities: first, we have a chance to influence the AI trajectory by steering critical conversations—such as those connecting climate justice, extractivism, and conflict in Africa. Second, we must recognize the power of collective resistance, movement-building, collaboration, and keeping community at the heart of all innovation.

AI can indeed accelerate MERL efforts, but this depends on rigorous benchmarking and thoughtful evaluation of the tools intended for use in Africa. Building a robust evidence base of successful use cases is essential to foster trust—an indispensable element for the widespread adoption of AI technologies.

Next steps

The next phase of this research involves disseminating a multilingual survey (in English, French, and Portuguese) to gather further insights from the NLP-CoP. We will also conduct more in-depth interviews and a comprehensive literature review.

We thank all participants for their contributions and invite ongoing collaboration as we finalize the landscape report. Feedback on the preliminary findings can still be shared through the Google Form here.