As artificial intelligence (AI) transforms higher education globally – from intelligent tutoring systems to strategic institutional forecasting – universities in developing countries stand at a critical crossroads. In the buzz around AI, an essential prerequisite – robust data governance and data quality – is often overlooked. Universities possess large volumes of data in various forms, including student, faculty, campus and marketing data, yet they often lack the capability to unlock value from the data sources that can enhance the student experience and overall campus efficiency. For institutions in low- and middle-income nations, focusing on data quality and management is vital to unlocking AI’s potential and ensuring that equity gaps with advanced countries do not widen.
The promise of AI in higher education
AI offers transformative opportunities for higher education, such as personalised learning and student services, where universities can identify at-risk students early, predict enrolment trends, optimise resource allocation, and boost retention rates and overall institutional efficiency. AI can enhance learning outcomes by about 20 per cent when effectively implemented, according to estimates from a 2024 World Economic Forum report. McKinsey & Company have further suggested that AI can reduce administrative workloads by 30 to 40 per cent, allowing educators and administrators in resource-constrained environments to focus more effectively on high-value tasks.
- The future of liberal arts may belong to the Global South
- Transforming mental health in South African universities: the promise of AI chatbots
- Data management strategy in higher education: a blueprint for excellence
Yet, despite this promise, the adoption of AI remains uneven across developing countries around the world, primarily because of significant data readiness challenges.
Data challenges for universities in developing countries
Universities in developing nations frequently encounter key data governance issues when considering how best to implement AI solutions within their institutions. AI readiness is often held back by:
- Fragmentation of data: Student, faculty, research and financial data often exist in siloed systems, hindering comprehensive data analysis and AI effectiveness
- Inconsistent data standards: Different departments might use varying definitions for crucial terms such as “graduate”, “enrolment” or “research output”, resulting in inconsistencies and unreliable analytics
- Manual data processes and outdated infrastructure: Reliance on manual data entry and legacy systems increases the risk of errors, shadow data and data reliability
- Lack of clear data oversight: Many institutions do not clearly define roles or assign responsibility for data governance, creating accountability gaps.
About two-thirds (64 per cent) of institutions in developing countries lack formal data governance structures, significantly restricting the effective and responsible implementation of AI solutions, according to a 2020 Unesco study.
Research evidence and policy insights
Countries with robust higher education data quality and interoperability have a 35 per cent higher rate of successful AI deployment, according to the World Bank’s EdTech Readiness Index (2022). The OECD’s 2023 Education Policy Outlook further identifies robust data governance as essential for equitable AI integration in higher education.
A 2022 Gartner survey reinforces this view, reporting that 87 per cent of university leaders recognise that addressing data silos and clearly defining data ownership are prerequisites for scaling AI effectively.
Strategies for building essential data foundations for universities in low- and middle-income countries
Universities should make a bold strategic commitment to data and AI with executive-level sponsorship, strong business cases and dedicated investment. This commitment, which starts with effective data governance, includes rules, policies, roles and processes designed to ensure data quality, accessibility, privacy and integration. Universities should therefore prioritise:
1. Standardising data definitions
Consistent data definitions across academic departments and colleges within universities are crucial. Unified definitions ensure accurate and actionable AI-driven insights. Universities should develop and implement clear data definition guidelines to achieve data consistency. A practical user case would be that often a significant portion of faculty research is stored in print-only journals or PDF documents without metadata or DOI indexing.
2. Improving data quality
Institutions must commit to regular data validation, cleansing and routine audits to ensure that data sets feeding AI systems are accurate and reliable. Universities should invest in automated data validation and cleansing pipelines. A useful model exists at UC Berkeley, which implements validation via Apache NiFi and Airflow to ensure accurate student performance records feeding into recommendation systems.
3. Building system interoperability
Integration of core institutional systems, such as student information systems, learning management systems, human resources and research databases, is critical across universities. Strategic investments in application programming interfaces (APIs) and data warehouses facilitate real-time data sharing, essential for predictive analytics and institutional strategic decision-making.
4. Ethical data management
Institutions must align with ethical frameworks such as Unesco’s 2021 recommendation on AI ethics, emphasising fairness, transparency, accountability and privacy, to avoid amplifying biases or disparities through AI systems. A useful model could be the University of Edinburgh, which established an internal AI ethics committee, GDPR-aligned data protection standards and comprehensive data literacy programmes for its community. Bias in data sets refers to systematic and unfair distortion in data that leads AI systems to make prejudiced or inaccurate decisions. This can arise during data collection, labelling, processing or even in algorithm design. Carnegie Mellon University runs fairness audits on data sets used in admissions AI to prevent gender or socio-economic bias in student selection. Universities should define what constitutes unfair bias and then regularly test data sets and models for bias.
5. Clear data stewardship and governance structures
Establishing clear roles and responsibilities, including data governance boards and data stewards within institutions, ensures accountability, oversight and sustainable data governance practices. Arizona State University (ASU) undertook extensive multi-year data quality and governance initiatives, enabling successful predictive analytics to enhance student outcomes. Like Edinburgh, ASU’s approach illustrates practical pathways for developing country institutions to build robust, effective data governance frameworks to support AI integration.
Increasingly universities have a chief data officer or AI ethics/data governance committee that define policies for data ownership, accountability, usage rights and quality standards. This approach creates clear accountability and ownership for data quality within universities.
Data-driven foundations first
For universities in developing countries, the potential benefits of AI are vast, yet realising these benefits requires foundational investments in data governance and quality. Institutions must recognise that effective AI solutions begin with reliable, well-governed and ethically managed data.
By focusing on data quality, stewardship and governance, universities in developing nations can harness AI’s potential to drive institutional growth, improve educational outcomes and foster equity in higher education. Ultimately, in the rapidly evolving AI landscape, quality data, not sophisticated algorithms, will distinguish successful institutions.
Cameron Mirza is chief of party at the International Research and Exchanges Board.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
comment