Enabling Enterprise AI Adoption Through Next-Generation Governance

Realizing AI’s opportunities, measuring and managing both benefits and risk, becomes difficult without appropriate and effective guardrails in place. Such guardrails, more commonly referred to as governance, are critical — not to slow down AI adoption and implementation, but rather to accelerate and optimize its responsible and effective use throughout the enterprise while creating new opportunities.

By implementing effective governance frameworks and policies, organizations can efficiently recognize, comprehend and measure the potential risks and opportunities associated with AI. This enables them to align their governance practices with the specific use cases they have identified and agreed upon, and ultimately make well-informed decisions regarding the utilization of AI.

In this paper, we share some best practices for aligning AI efforts with the foundational elements of a strong AI governance program, starting with forming an advisory board and defining use cases.

First things first — select an advisory board and use cases

Most organizations are familiar with governance. They already have policies in place for various technology solutions that can serve as a guide for effective AI governance. Additionally, the lessons learned from establishing governance for other business processes and technologies can expedite the development of an efficient AI governance approach and framework.

The foundation for AI governance begins with people who, in this case, will form an advisory board that should guide and promote the adoption of this rapidly evolving technology to meet business needs. The advisory board typically includes senior managers, legal counsel, compliance managers and experts from across the organization who understand the business’s defined goals and the importance of driving transformation.

The role of an advisory board

An advisory board, which might also be called a steering committee, is primarily responsible for providing strategic direction and oversight and ensuring alignment with the organization’s broader mission and objectives. The advisory board should define the vision and objectives for deploying AI in the organization and outline the opportunities and risks.

Ultimately, the advisory board is responsible for identifying the key areas where AI can be beneficial, and it promotes adherence to ethical guidelines and regulatory compliance. It also ensures that the aspects of the AI governance framework align with legal requirements, industry standards (for example, NIST AI RMF controls), the company’s ethics, and moral principles such as transparency, fairness and security.

It is critical for the advisory board to represent leaders and experts throughout the organization to ensure a multidisciplinary approach and mindset. It should be comprised of key stakeholders, including those that have knowledge and expertise in key areas, such as AI technology, internal audit, cybersecurity, compliance, legal counsel and business operations.

Once the advisory board is identified, its first order of business should be to identify and evaluate potential AI use cases for the organization.

Use cases pave the way to identifying governance requirements

One of the first and most essential functions of the advisory board is identifying and examining individual use cases for AI in the organization, ensuring that each use case gets customized controls based on its unique needs while considering the potential benefits against any risks that need to be addressed and managed.

As we noted earlier, governance is not a new concept for organizations. Most have a high level of familiarity and often sophisticated governance policies and practices in place around numerous functions — ERP systems, supplier management, cybersecurity and more. Most organizations likely have detailed governance frameworks that they apply to these functions, ones that may have taken months or even years to refine.

When it comes to AI, the key points here are that (1) there is no need to reinvent the wheel with a unique approach to AI governance — it’s likely existing policies and practices offer some nuances and specificity that can be used as foundation components for AI governance and (2) the governance components and controls should be flexible and adapt to how AI will be used — i.e., the use case.

Each use case may have unique capabilities and requirements that may require different controls in the governance framework. For example, an AI use case of identifying preventive maintenance needs will carry a much different set of risks than a use case that supports hiring decisions. It is also possible that some controls defined in the organization’s current governance programs, policies and frameworks may not be necessary for a particular AI use case. Which controls are and are not relevant and necessary depends on the specific AI use case and its context. For example, an AI deployment that is considered a closed system, where no queries or data extend beyond the internal systems, may not need stringent privacy or data sovereignty controls.

The advisory board should support the examination of the implications of the use case and evaluate its viability. This includes:

• Identifying the problem the AI solution will solve and setting measurable objectives for success.
• Ensuring visibility into the quality and relevance of the data sources used.
• Determining data requirements such as storage, processing, privacy and security.
• Determining the high-level technical components needed for implementing the use case.
• Assessing compliance requirements, including data protection laws and industry regulations.
• Addressing ethical concerns and incorporating principles of fairness, transparency and accountability.
• Performing a risk assessment to understand potential challenges and unintended consequences.
• Identifying and engaging stakeholders affected by the use case.
• Comparing potential benefits against costs to justify investment in the use case.
• Creating metrics to evaluate the success or failure of the use case.

Once these tasks are completed, the next step is to outline further the use case to determine how it fits into a broader governance framework.

Refining a use case requires incorporating considerations that are typically defined in an AI governance framework. These considerations will help accelerate the development of the use case and make the deployment align with business goals.

Keep in mind that different use cases may have different considerations. However, most use cases share some common principles that are encapsulated in the governance framework. These principles, summarized below, can speed the development of use cases while also preventing essential elements from falling through the cracks.

Aligning governance to use cases

Now that we’ve set up the advisory board and defined the AI use cases, it’s time to discuss governance.

With the introduction of any innovative technology, there are considerations that must be addressed to help drive success. For example, a benefit versus risk analysis is a critical element for vetting any use case. In addition to defining the opportunities created by implementing an AI use case, it is important to identify and manage risk properly to realize the full potential offered by that use case. Only then can the specifics of the use case be aligned with the components of a governance framework.

In many cases, particular components of a governance framework may not be applicable or may have reduced importance during the alignment process. Even so, there are some basic foundational tasks that are common across most use cases:

Understanding and communicating the benefits of governance

One of the biggest benefits offered by a strong governance program is ensuring AI adoption is aligned with organizational strategy and designed to maximize business value.

The benefits offered by a properly executed governance program should be shared with the core stakeholders that are impacted by the defined use case and, eventually, the broader enterprise. Some of these benefits include:

Shaping AI governance: internal and external dynamics

Governments, legislative bodies and other authorities are working to address concerns related to the adoption of AI technologies and are developing or modifying legislation, guidelines and compliance requirements to meet the challenges posed by AI. Organizations that are adapting their governance programs for AI use cases need to consider how external factors will influence the design and application of their frameworks. Whether an organization is modifying existing frameworks or creating new ones, it is important for them to conduct their own due diligence.

Selecting the appropriate governance policies depends on the specific needs of the organization. For example, certain businesses may have charters, codes of conduct or foundational perspectives that influence the use of AI technologies and guide appropriate use policies. Therefore, the organization needs to consider and define the following factors:

As noted earlier, an important aspect of AI governance is establishing company standards for the implementation and use of the technology, and how those standards impact use cases. Fortunately, many organizations already have broader technology-related standards in place that can be used to support the development of AI governance standards. This simplifies the process of creating specific AI governance standards.

Call to action for AI governance

Identifying appropriate governance for an AI solution is one of the most critical steps for an effective deployment of the technology. However, determining the applicable policies and components of governance is a task that should not be taken lightly. Assembling the appropriate controls to protect data and ensure ethical use of AI systems will take a team effort, where stakeholders as well as management have a clear understanding of the goals being pursued. Obviously, there are steps to instantiating appropriate governance before deploying any AI solution, which include the following:

• Create an advisory board.
• Align the governance framework, policies, standards and controls.
• Educate users.
• Develop an inventory of existing AI technologies and use cases and rate those use cases for risk.
• Apply controls as appropriate to the use cases, including red-team testing of high-risk use cases.

Of course, AI systems evolve over time and require measurement for effective long-term management. Governance goes a long way toward providing the foundation that supports an AI deployment, while also setting the stage for reiteration and agility.