Responsible AI structure
During the DataIQ 100 Summit, former Chevron CDO, Ellen Nielsen, presented a four-prong structure for addressing responsible AI which she encouraged data leaders to adopt:
- Principles
- Assessment
- Governance
- Communications
Principles
When it comes to focusing on principles, Nielsen encouraged data leaders and their respective companies to consider their risk profile and what guidelines they should provide as key principles for any AI projects.
As an example of what they can look to for inspiration, Nielsen highlighted Microsoft’s AI principles as a useful framework. Microsoft outlines six key principles for responsible AI, built around the guiding principles of ethics and explainability: accountability, inclusiveness, reliability and safety, fairness, transparency, and privacy and security.
Of course, not all businesses will be working in the same way as Microsoft and the products and services being provided will be vastly different, but the basic principles for responsible success should remain the same no matter what. This is why data leaders need to work closely with other leaders and decision makers within the business to examine what the risk profile and guidelines for their needs should be.
Assessment
In her presentation, Nielsen highlighted the importance of impact assessments with the desire to ensure the organisation has an inventory of AI projects. Organisation is key to responsible AI and within that, each tool and project needs to be assessed regularly and thoroughly.
Currently, the regulations surrounding AI are vast, varied, and often vague, so it is essential that data leaders instil a routine of assessment and critique early on to be able to handle incoming regulatory changes. It seems inevitable that regulations will change and be implemented in the next couple of years, so businesses need to prepare for this and be flexible, robust, and ready to meet the new changes.
Governance
Assessment leads neatly into governance. Here, Neilsen encouraged companies to ensure there was a human in the loop. It is important to remember that AI is a tool, not necessarily a solution, and it often takes a human to spot errors and inconsistencies. As AI continues to evolve, the guiding principle of being human-led must remain the same. Think of humans as the pilot – the aeroplane is the tool, but it still needs a human to operate it effectively, safely, and responsibly.
Additionally, Nielsen explained that data leaders need to ensure AI maturity assessments were regularly completed. The reason for this is it is important to view AI over its life cycle with regular checks to assess current state. Maturity assessments are a process designed to evaluate the current AI capabilities of different businesses, identify gaps and areas for improvement within those businesses, and develop a roadmap.
Nielsen championed the audience to view the assessments from the Responsible AI Institute and the need for oversight of AI from a high-level board. DataIQ members have previously discussed this issue at roundtable events and the learnings can be found here.
Communications
Finally, Nielsen discussed the importance of communications in her four-prong structure. Neilsen stressed the importance of regular compliance training for AI and the need for strong policies as these will develop the data culture of the business and ensure a higher chance of success in the ever-changing field of regulations.
For an example of what to look for with communication, Nielsen recommended the NIST website and frameworks as a useful entry point for the metrics companies should be using.
Communication internally and externally is key to driving success as it ensures all existing hires and customers are on board with the journey, as well as bringing any new customers and hires up to speed rapidly.
There is a level of distrust with AI tools, and this is often down to people not fully understanding or appreciating the use of AI. With improved communications and installing roles such as data owners and data stewards, businesses can swiftly and effectively dispel any myths or misconceptions about AI and its role within an organisation.
Scotiabank received the title for the Best Responsible AI Programme at the 2024 DataIQ AI Awards. The work Scotiabank has put into its AI programme was noted for its ethics and responsibility in adopting a public data ethics commitment statement and creating and implementing tools.
The 2024 DataIQ 100 Summit provided incredible insights into how leading data professionals have approached a multitude of topics. It also provided a face-to-face opportunity for attendees to learn from, connect with, and develop lasting relationships with some of the most prominent minds in the industry.
To register for the 2025 DataIQ 100 Summit, click here.
