Learning to Recommend: AI, Human Counselors, and Job Matching in Kenya

This project investigates the potential of AI to improve the efficiency and effectiveness of job matching in low- and middle-income countries. Partnering with two innovative NGOs in Kenya—Tabiya and Swahilipot—the researchers evaluate “Compass,” an AI-powered career counseling tool that helps job seekers identify skills from past experience, create employer-friendly CVs, and receive tailored job recommendations. Compass can operate independently or alongside human counselors, offering scalable support in contexts where caseworker capacity is limited. Through a randomized evaluation with one control group and three treatment arms (human-only, AI-only, human + AI), the researchers measure impacts on job search behavior and downstream outcomes, as well as operational productivity. This design enables them to answer the critical question: how do AI tools complement versus substitute for human expertise in career guidance? Researchers will cross-randomize information (“feedback”) received by both humans and AI to study how each learns about job seekers’ preferences and how that learning shapes advice quality. This research has significant practical value for employment platforms and public employment services seeking to improve labor market efficiency in contexts with low job seeker support capacity. By better surfacing informal skills, Compass especially benefits marginalized job seekers – youth, women, and those without formal education – opening pathways to formal employment and improved labor market inclusion.