AI Governance in the Public Sector: Why Data Governance Must Lead the Way

Drawing from extensive conversations with data professionals across government departments, local authorities, and NHS trusts, I’ve identified a critical gap in how many organisations, including in the public sector, are approaching AI implementation. The solution isn’t about creating yet another new governance structure, it’s about evolving your existing data governance framework to encompass AI.

Understanding the Gen AI Shift

Generative AI represents a fundamental evolution from traditional predictive systems. As AI expert Lara Gureje explained, when I interviewed her on my podcast, generative AI is “AI on steroids”, creating entirely new content rather than simply analysing historical patterns. In public sector contexts, this means systems that can automatically generate benefit decision letters, policy recommendations, and citizen communications at unprecedented speed.

This creative capability introduces governance challenges that traditional AI oversight wasn’t designed to handle, particularly when decisions directly impact citizens’ rights and access to services.

The Fragmentation Problem

Through my work with clients across many sectors, I have observed many organisations implementing AI governance as a separate function from data governance. This fragmented approach creates challenges. Organisations rush toward exciting AI implementations while overlooking fundamental questions: Do we have quality data fit for AI purposes? Everyone wants to do the fun stuff but not put the foundations in place first. Understood and good quality data serves as the foundational raw material for AI models. Without solid foundations, trust diminishes, and analytics become meaningless.

Adding a separate approach for AI governance just adds another layer of complexity for those trying to implement AI solutions. Consider a policy officer implementing AI for benefits assessment, navigating numerous separate processes for ethics committees, GDPR assessments, data governance assessment, and data quality reviews! This fragmentation both creates barriers to innovation while increasing oversight gaps.

Why Data Quality Demands Priority in Government

Public Sector AI applications carry unique risks. Poor data quality doesn’t just reduce efficiency; it could result in all manner of areas such as:

•            Incorrect benefit calculations affecting vulnerable families

•            Biased decision-making in social services

•            Ineffective healthcare resource allocation

•            Unfair immigration and justice outcomes

To mitigate against such things happening we need to make sure that the data the AI models are being trained on is good enough quality. The “rubbish in, rubbish out” principle becomes a matter of democratic accountability, where data quality failures become governance failures affecting real people’s lives.  It’s not all about the exciting new technology, Sol Rashidi in her book Your AI Survival Guide states that successful AI deployment depends 70 percent on governance, strategy, and human factors, not technology.

Six Principles for Integrated AI Governance

So you may be wondering how best to implement useful AI governance at your organisation?  Well, I recommend these essential strategies for successful implementing and integrating AI governance with your existing data governance approach:

•            Evolve existing data governance frameworks rather than creating separate AI oversight structures

•            Ensure data quality and governance maturity before AI

•            Establish single points of responsibility spanning both data quality and AI model outcomes

•            Document AI system purposes, data sources, and decision logic

•            Apply risk proportionate controls matching the potential impact on citizens and public trust

•            Develop AI literacy within existing data governance teams rather than creating separate expertise silos

The Integration Advantage

Integrating AI governance within data governance frameworks delivers measurable benefits across three key areas: Operational simplicity means teams navigate one governance process rather than multiple separate frameworks, reducing complexity and improving compliance rates. Resource efficiency allows organisations to leverage existing capabilities, roles, and processes rather than duplicating efforts across separate governance structures. Comprehensive coverage ensures data quality, ethics, compliance, and operational considerations receive coordinated attention, reducing oversight gaps.

Moving Forward: Creating Sustainable Change

Success requires making compliance straightforward rather than burdensome. Integrated AI governance should streamline decision-making while ensuring comprehensive oversight of citizen-impacting systems.

Generative AI is growing so quickly that without early governance foundations, it could result in disaster. This urgency demands immediate action building on existing your data governance framework.

Effective AI governance isn’t about perfect technical oversight; it’s about creating sustainable organizational practices that ensure AI serves citizens while maintaining democratic accountability and public trust.

The choice facing public sector organisations is clear: integrate AI governance into proven data governance frameworks now, or risk fragmented approaches that compromise both innovation and citizen protection. The public deserves governance structures that deliver both technological advancement and democratic oversight.