Overview of CSI spend on social and community development in 2025

[CASE STUDY] Gender-responsive monitoring and evaluation in feminist tech spaces 

For organisations working at the intersection of gender, technology and justice, measuring impact can be challenging. Systemic, long-term goals such as trust, capacity-building and access to justice are not easily quantifiable. Yet rigorous, gender-responsive monitoring and evaluation (M&E) is essential to ensure that programmes respond meaningfully to lived experiences, particularly in contexts marked by inequality, violence and exclusion. 

Gender Rights in Tech (GRIT) is a nonprofit advancing a feminist M&E approach through its suite of digital tools designed for survivors of gender-based violence (GBV). GRIT’s tools are co-created with those most impacted by GBV and build on feminist evaluation traditions that ask deeper questions about power and justice.  

These tools include: 

• The free GRIT mobile application, which is data-free and supports over 12 000 users with features such as a panic button linked to rapid response services and a secure vault for storing audio, visual and narrative evidence.  
• Zuzi, a generative artificial intelligence (AI) chatbot designed to provide trauma-informed, localised support to GBV survivors, especially those underserved by mainstream systems, including queer youth, rural communities, migrant women and sex workers. 

Zuzi reflects how people in South Africa actually speak about harm and violence, and how they ask for help. “It uses accessible language, incorporating culturally grounded concepts of justice, community and survival,” explains Leonora Tima, founder and managing director of GRIT.  

GRIT’s approach to AI is shaped by a gendered African lens that disrupts the power dynamics underpinning most AI systems, which primarily cater to people who are white, highly literate and tech-savvy. “It’s not just the language that’s Eurocentric – it’s the entire architecture: who AI is built for, what it assumes justice looks like and what kinds of knowledge it considers valid,” says Tima.  

This means designing for voice-to-text users, shared phones, low data and layered trauma, with context (rural, queer, multilingual, working class) as the uppermost priority. 

“We’ve had to hold the tension between protecting data and not forcing people to repeat their stories. Our system is designed to minimise that burden while still giving us the insight we need to make Zuzi better. This is where our feminist approach to M&E shows up: we treat evaluation not as surveillance, but as care. Not as extraction, but as learning. Disrupting AI isn’t just about clever prompts or language models. It is about who we design for, who we listen to and what we count as success.” 

Reimagining how impact is measured 

GRIT’s M&E processes don’t just measure outputs – they question assumptions. “We interrogate what ‘effective’ means and whether these tools expand access or just replicate formal, legalistic pathways that many survivors can’t or won’t use,” according to the GRIT team.  

“Our work with survivors has shown that AI systems often erase the nuance of African experiences of violence, community and justice or – worse – encode biases that stigmatise them.” GRIT measures Zuzi’s impact not just by how many people use it, but by how they experience the chatbot, and then builds metrics that can speak to what Zuzi reveals about power, trust and care.  

“Our M&E approach is rooted in feminist principles, which means we don’t chase vanity metrics,” says Tima. “Instead, we ask: how do survivors navigate support differently when given a tool that actually listens, responds in their language and doesn’t make them start from scratch every time, or make assumptions about their access?” 

Much of GRIT’s M&E framework grew directly out of its WhatsApp line, where it has supported survivors in real time.  

“From those interactions, it became immediately clear that the way someone moves through justice or care-seeking processes is deeply shaped by their race, gender, class, sexual orientation, language, socioeconomic circumstances and how those intersect with the institutions they’re asked to engage with when seeking help,” says Tima.  

This means redefining what ‘progress’ means. “Sometimes it looks like asking for a protection order. Other times, it’s finally talking to a family member after years of silence. It might mean using a healthier coping mechanism, even if the broader legal system still fails them,” according to the GRIT team.  

GRIT reflected on how traditional M&E metrics fail to capture the weight of trauma or non-linear recovery. Therefore, GRIT’s evaluations track aspects like confidence in Zuzi – not just by what survivors say they’ll use it for, but by how they actually use it compared to other options like Google or helplines.  

“We look at emotional tone, drop-off points, what gets asked repeatedly and where people go silent. We design our indicators with these aspects in mind as they tell us as much as the data we collect from the content of the conversations themselves,” says Tima.