Evidence Generation
Asia
Evidence Generation
Asia
Funded by IPA’s Consumer Protection Research Initiative and in partnership with IPA Uganda, Center for Technology Disputes Resolution - Uganda (CTDR-U), and Boda AI Factory, researchers are conducting a pilot in Uganda to assess whether an AI-powered legal companion can be an effective and scalable tool to improve consumer redress.
Digital financial services have increased financial inclusion for millions of people who were previously excluded from the formal financial system. However, this has been accompanied by rising consumer protection challenges—including fraud, unauthorized fees, agent misconduct, and deceptive practices—and users have difficulty with the redress process when problems arise. For instance, in Uganda, an IPA survey found that more than 75 percent of digital financial services users reported experiencing at least one problem in the previous year, yet most never took action or resolved their complaint.1 A key driver of this gap is information asymmetry: many users do not know their rights, do not understand complaint procedures, or distrust redress systems enough to act.2
Previous research in Uganda found that structured legal aid significantly improved complaint resolution rates and trust in digital financial services.3 However, this program, which was led by case workers, was determined to be too costly to scale. While some cases required more legal support, a sizeable portion were recurring with known solutions, making them ripe for automated solutions. Providing users with low-cost consumer redress avenues offers the opportunity for innovative, technology-driven solutions such as AI support. AI systems are increasingly used across sectors to improve information access and knowledge for populations in low-income countries. Yet evidence is lacking on its effectiveness in the legal sector. Whether AI-powered tools can provide consumers with effective, low-cost support to navigate redress remains an open question.
Wakilibot is an AI chatbot that helps Ugandan digital finance users identify problems, understand their consumer rights, submit complaints, and escalate disputes when needed. It is designed for users with limited digital literacy and combines rules-based logic with retrieval-augmented generation (RAG) AI to deliver accurate, context-sensitive responses in English and local languages. WakiliBot will be accessible via the web, SMS, USSD, and voice channels, making it available to both smartphone and feature phone users.
In partnership with IPA Uganda, CTDR-U, and Boda AI Factory, researchers are conducting a pilot study to assess whether WakiliBot can be an effective and scalable tool to improve consumer redress.
Over six phases, researchers will design, test, and refine WakiliBot. To evaluate WakiliBot’s system performance and user learning, researchers will measure the accuracy of its responses against lawyer-generated standards. They will also collect cost data across the project to assess its potential cost-effectiveness relative to human-provided legal aid.
The pilot will start with 100 to 150 users in user acceptance testing, expanding to 1,000 to 1,500 in a soft launch, and ultimately reaching up to 10,000 users if performance benchmarks are met.
Results will be available in 2027 and will inform a full impact evaluation.
1. Mazer, R. & Bird, M. (2021). Consumer Protection Survey of Digital Finance Users: Uganda", https://doi.org/10.7910/DVN/ROLCU4, Harvard Dataverse, V1
2.,3. Chemin, Matthieu, and Allan Katimbo C. Katurebe. "The Causal Effect of Dispute Resolution Mechanisms on Economic Development: Evidence from a Randomized Experiment." Available at SSRN 6596558.
