As the world of AI continues to expand, organisations are being called to navigate a challenging balance between innovation and governance. Next month, at The Data, AI & Analytics Conference Europe 2024, Deep Lidder, Data Governance Delivery Lead at Cervello, a Kearney company, will take the stage to share his expertise on AI governance. Lidder’s session, scheduled for Wednesday, October 16, 2024, from 3:00 PM to 3:40 PM, will cover key pillars for organisations to consider when building a comprehensive AI governance framework. Ahead of the event, we caught up with Deep to discuss some of the core themes of his session, from key pillars of AI governance to real-world challenges in AI deployment.
Your session focuses on AI Governance. Could you share some of the key pillars that organisations should consider when establishing a comprehensive AI governance framework?
“When designing and implementing an AI governance framework, organisations should consider several key pillars. Fairness is essential to prevent AI models from perpetuating bias or discrimination, particularly in sensitive deployments such as hiring or healthcare. Transparency is critical for building trust, as it allows stakeholders to understand how AI systems make decisions and which data they use. Accountability is equally important, ensuring that roles and responsibilities are clearly defined, so that if there are any errors or unintended consequences, there is a clear line of responsibility. Explainability also supports transparency by ensuring that we have understandable explanations for AI outputs, particularly important in regulated industries or high-risk settings.
In addition, organisations should integrate risk management practices to measure, monitor and mitigate potential operational risks associated with AI. Data privacy and security are also important, particularly given regulations around data protection. Without these considerations, there is a risk of negative consequences, including customer dissatisfaction, reputational harm, or non-compliance with regulatory requirements.”
In your experience, what role does senior leadership play in shaping an organisation’s AI governance strategy, and how can they effectively balance innovation with responsible AI use?
“Senior leadership plays a critical role in shaping an organisation’s AI governance strategy, setting both the tone and direction. By prioritising AI governance, leaders can ensure it becomes an embedded part of the organisation’s culture and values. One of the most significant contributions of leadership is communicating the importance of AI governance across all levels of the business and ensuring that it receives the necessary resources for effective implementation.
Balancing innovation with responsible AI use should not be viewed as a conflict; rather, governance acts as an enabler of innovation by providing the guardrails within which your data scientists and developers can confidently and safely experiment. Senior leadership also have the ability to drive cross-functional collaboration to share learnings and align on key decisions. Support for data and AI education is essential, helping all employees to understand the risks associated with AI, and make informed decisions on how these can be effectively mitigated.”
AI regulation is a growing concern across industries. How can organisations stay compliant with evolving regulations without stifling AI innovation?
“Regulations are evolving, and maintaining compliance requires a proactive and flexible approach. Organisations should regularly monitor the regulatory landscape and update their approaches to AI governance accordingly. For example, the UK has signaled taking a sector-specific approach (‘A Pro-Innovation Approach to AI Regulation’ 2023), with existing industry regulation providing the foundation for AI. Organisations should take the opportunity to proactively engage and communicate with relevant regulators (e.g. Ofgem, FCA) to input into discussions and reviews as regulation continues to develop.”
Could you provide an example of a real-world challenge you encountered when moving from AI ideation to deployment, and how you overcame it?
“A common challenge that organisations face when moving from AI ideation to deployment is managing data quality, especially when attempting to scale AI models. In a recent project involving a marketing use case, we encountered significant data quality issues as we progressed from the ideation phase to full deployment. During the initial phase, we were working with a smaller, more controlled dataset, which produced strong results. However, as we expanded towards operationalisation, the complexity increased dramatically due to the integration of data from multiple systems with varying formats and standards. This created performance issues in the AI model, as it struggled to process inconsistent and incomplete data.
We overcame these challenges by designing and implementing strong data governance and data quality practices. This included automating data validation checks to ensure that errors and inconsistencies were detected in real-time and could be corrected before they impacted the model. We also worked closely with cross-functional teams to establish clear ownership and responsibility for data quality across the organisation. Defining company-wide quality standards helped ensure consistency, while ongoing monitoring of data inputs post-deployment allowed us to detect issues and maintain quality outputs at scale. By collaborating closely with business and technical teams, we were able to operationalise the AI model successfully.”
Ethical AI is a central theme of your session. What steps should organisations take to ensure their AI systems are not only compliant but also ethically sound?
“Organisations should identify and define clear ethical guidelines that align with the organisational values, mission and regulatory requirements. There is also a need to drive data and AI education so that stakeholders across the organisation understand the different types of ethical considerations, and how these can be embedded as part of the AI development process from the outset. A dedicated team should review and provide oversight for AI projects from an ethical standpoint. A key aspect of this approach is involving diverse, multi-disciplinary teams throughout the AI lifecycle, which helps to ensure that a range of perspectives are considered. By prioritising diverse inputs and implementing ethical controls, organisations can ensure their AI systems not only meet regulatory requirements but also align to ethical standards.”
There’s still time to register for The Data, AI & Analytics Conference Europe 2024 and hear more from Deep Lidder and other industry leaders. Don’t miss this opportunity to dive deep into the future of AI governance. View Tickets Here!
