Lisa Udechukwu works in a part of the AI industry most people never hear about: the quiet, unglamorous work of making sure the data that trains intelligent systems is actually good. As a data quality analyst focused on AI annotation, data integrity, and trustworthy AI evaluation, she has a front-row seat to why so many AI tools fall short for users outside the Western world, and her diagnosis is straightforward. The problem is not the technology. It is the data.
Her argument cuts to something that people who use AI tools in Lagos, Nairobi, or Accra have likely already felt in practice. When a voice assistant struggles with an accent, when a translation tool fumbles a local language, when a content recommendation feels completely out of step with your reality, that is not a glitch. It is a reflection of what the system was, and was not, trained on.
Data quality, annotation accuracy, and whose experiences get included in training sets are decisions made long before a product ever reaches a user. Udechukwu's focus on trustworthy AI evaluation puts her at the center of those decisions, which makes her perspective worth following closely as the conversation around AI fairness and inclusion continues to grow.
Originally published by BusinessDay.