Keeping Your Data Governance in good shape and Up to Speed to meet the surge in AI usage
Artificial intelligence (AI) is in the midst of a massive hype cycle at the moment. Everyone in our space is talking about it much more than usual inspite of the fact most have been using it for years but were calling it data analytics or data science.
Chat GPT and its myriad of look-a-likes have created a spark of enthusiasm and interest in the amazing potential that data has to offer.
This means there’s been a bit of a wave of new AI initiatives and ideation happening at the coal face looking for ways to prove the value of AI.
A major challenge however is hitting people head on – the state of their data and systems. AI needs good data, in good shape to work, and most businesses have underinvested over the years in their core underlying architectures and processes required to build and sustain a high performing data environment.
Now as everyone is clamouring to do more with their data using more sophisticated methods such as Generative AI they are realising that they are not ready.
The data is still hampered by legacy systems issues and the processes around managing that data are often struggling to cope.
A major challenge I see businesses facing is in adapting their existing data governance and risk frameworks to ensure they are robust enough to manage AI's unique complexities.
With the applications of AI often resulting in a heavily automated set of decisions or outcomes, the need for controls and checks and balances to manage the associated risks mean that Data Governance has shifted from a nice to have to a must have.
A lot of our clients have however already invested heavily in data governance and are feeling the pinch somewhat as they realise the requirements have suddenly picked up pace.
We’ve been helping our clients assess the changes and figure out how to adapt and extend their existing frameworks to stay ahead in the AI landscape.
This blog plays back some of these learnings and advice for you to consider in your own business.
The Challenges of AI Governance
Because AI uses existing data to learn and its dynamic nature, its ability to not just amplify human biases, but unwittingly bake them into decision making criteria means our clients have suddenly faced into significant governance challenges.
Their applications, be they recruitment tools or customer service based deal with real human beings and as such there is no wriggle room for the “computer to say no” at an inopportune moment.
AI governance comes into play to ensure the business is creating the right policies, assigning suitable decision rights, and ensuring organisational accountability for the risks and decisions associated with AI.
AI Governance is complex because of AI’s predictive and generative capabilities, which add layers of ambiguity and require careful management to balance the value it delivers and the risk of it getting things wrong.
Why Existing Governance Frameworks Need Extension
While many of our clients have established or are starting to establish strong data governance frameworks, these now need to be extended to accommodate the AI-specific considerations.
The potential human and societal impact that AI can have means we all need to put in place additional governance measures to ensure transparency, accountability, and ethical usage.
Here at Beyond we have been advising our clients on how they should incorporate these AI challenges into their own Governance Frameworks and what follows is a summary of what we have been suggesting:
Key Steps to Extend Your Governance Framework for AI
Adopt a Common Governance Framework
Some of our clients have approached Governance in an organic way over time, which whilst still effective means that the way its designed and communicated can become confused.
With AI Governance looming it makes for a good time to adopt a more formal or structured framework to how you think about Governance. There’s plenty of these around or available online, but we have described the framework Gartner suggest as a pretty good starting place that is going to do what it says on the tin.
This framework is useful as it groups up everything you need to get right into sensible and meaningful categories as below. Think of these as the table of contents for your Governance Framework, or a checklist that you have every base covered.
Section 1 - Mandate and Scope
Define and establish the responsibility, authority and support from senior leadership.
Provide focus on core AI principles, governance charter, and executive sponsorship.
Create a formal/signed off governance charter and agreement on relevant AI principles.
Section 2 - Structure and Roles
Define roles, responsibilities, and skills necessary for proper governance.
Define the decisions required and the responsibilities for making them on governance functions and deployment style (centralised, decentralised, or hybrid).
Involve the right stakeholder participants and identifying the upskilling required to enable all parties to be effective in the process.
Section 3 - Process
Develop standards, policies, and guidelines for AI operations.
Establish procedures for smooth AI progress (research, development through to implementation and maintenance) and risk mitigation checks and balances.
Encourage consistent policy management and innovation through the appropriate review structure and processes.
Section 4 - Decision Rights
Establish authority and accountability for business, technology, and ethical decisions.
Balance the various dimensions of AI governance (organisational, societal, customer, employee).
Define decision rights for technical and business experts.
Section 5 - Culture
Align governance with organisational values and bias mitigation.
Promote an AI-first culture, ethical AI, and bias mitigation guidelines.
Consider the diverse cultural norms and regulations.
Section 6 - Communication
Focus on trust, AI adoption, and AI literacy.
Communicate AI progress, policies, and guidelines clearly.
Promote transparency and stakeholder engagement.
Using a framework like this makes a lot of sense as it’s a practical foundation to which you can begin to map all the AI-specific challenges.
By doing this you are breaking the challenge down into more manageable and familiar concepts and ideally ensuring that you are merging existing thinking with the new AI requirements.
As you embark on the process, focus your AI governance efforts on the most critical AI use cases with clear accountability across business, technical, and ethical dimensions. i.e. pick the ones that you can clearly identify have the potential to carry greater risk.
Formulating stakeholder discussions to explore the value and risk of your various AI use cases will help your organisation prioritise as well as consider the most relevant emerging regulatory considerations and ensure you are building compliance into any of your most significant AI driven outcomes.
This process needs to encourage the involvement of a diverse group of stakeholders. We often see Governance being relegated to the back office and the business stakeholders keeping their distance on what can be perceived as a dry, unengaging topic.
In addition to engaging participants from the directly relevant functions such as security, risk, audit, procurement, and compliance you need to make sure that the end business users/stakeholder groups are fully engaged and involved.
This wide group of stakeholders needs to jointly decide on AI governance steps and get the balance right between Governance and Value.
Keeping tabs on AI developments and progress is an important consideration as the group of stakeholders required will become more diverse the broader the scope and progress of AI usage develops.
Working through the framework at a high level a useful checklist of the key points to make sure you are getting things right would include:
Mandate and Scope - Establish a clear AI governance charter with executive sponsorship. This needs clearly defined so people identify the significance of AI Governance rather than perhaps just another piece of ‘internal process’. This involves defining your AI principles around trust, transparency, and diversity, and setting the governance scope to support your AI strategy and vision. Your AI charter should outline the scope, authority, and mechanisms for governance, ensuring alignment with organisational goals.
Structure and Roles – Review or determine the AI governance functions and deployment styles (i.e. centralised, decentralised, hybrid) so they mirror your existing data and analytics (D&A) governance approach. This will require the involvement of a diverse set of stakeholders to think through and manage your AI-related business decisions and address new challenges. Keep in mind that your latest AI developments are possibly happening in parts of the business that haven’t traditionally got very involved in Data Governance.
Processes – Think how to develop the standards, policies, and guidelines to streamline AI implementation. Without these its likely that getting AI related initiatives through the corporate approval processes will become incredibly sticky as every stakeholder group hits the brakes because they aren’t sure of how their role will impact the outcomes and what risk that might pose. E.g. they’ll choose to stop something happening at all as they safer outcome. By ensuring you have clearly defined steps and that these are consistent (read familiar) with your D&A governance, and you have the relevant procedures in place to resolve dilemmas and managing any potential public-facing AI issues, will go a long way to calming the nerves. You will need to factor in regularly updating these policies to reflect your evolving AI landscape.
Decision Rights – To work effectively you will also need to assign clear authority and accountability for specific AI-related decisions, involving business, technology, and ethical experts. There is no room for opaqueness or uncertainty which leads to the brakes coming on again and progress stalling. Be sure to define decision rights for critical and non-critical AI content, ensuring clear accountability so everyone knows who is doing what.
Culture – AI means a lot of different things to different people. You will have both extremes from the likes of me, who will wonder what all the fuss is about as this is what we have been doing for the past 20 years, through to the deeply sceptical that find the whole AI thing pretty scary. Fostering an organisational culture that supports AI adoption is super important and it needs to recognise peoples' feelings and views. To start, make sure you are embracing AI principles that align closely with your corporate values (and be seen upholding these values) and implement clear bias mitigation guidelines to ensure fairness and inclusivity.
Communication - Promoting AI literacy and transparency is therefore a really important consideration and one that is often forgotten. Start by educating your teams by communicate all about AI’s probabilistic nature and what that means or implies, set stakeholder expectations about how the business plans to engage and explore where it will go with AI as well as identify where it will not. Provide regular updates on AI progress, share the learnings – good and bad - to build trust, demonstrate transparency in your actions and garner support. Establish feedback channels to capture user input and improve AI systems iteratively.
Conclusion
I hope this article has helped understanding of why keeping your data governance framework up to speed with AI is so important and how crucial it is to extend your existing governance structures and processes to address the AI-specific challenges.
By taking the opportunity to adopt a governance framework, focusing on critical use cases, involving diverse stakeholders, and fostering an AI-supportive culture, you help your business ensure its governance efforts remain relevant, effective and above all safe.
This proactive approach will keep you on top of the issues and challenges and help you balance AI's value with its inherent risks which in turn should support driving successful AI adoption and maintaining the integrity of your governance framework and ultimately your business.