AI Governance: Is controlling AI processes useful? By Aditya Abeysinghe

AI Governance: Is controlling AI processes useful? By Aditya Abeysinghe

AI Governance: Is controlling AI processes useful? By Aditya Abeysinghe


Aditya-AbeysingheDeploying an AI system includes several processes from gathering data to build AI models to testing the system to be deployed. Most businesses which deploy AI systems require a set of policies at each process of this deployment. These policies include which approach to be used, which tools and methods to be used etc. These set of processes, called AI Governance, is used to define how an AI system needs to be deployed while enhancing management of risk, cost, and scope.

What is required to achieve AI governance?

AI governance of an AI system is regulated by the organizational structure which monitors AI processes from data collection to deployment. The structure may differ from the size of the system, number of teams involved, and the cost and time of completing the AI system. Usually, the structure has members with managerial and technical capacity and managers regulate policies. Therefore, a business structure with members who can handle an AI system according to both technical and business regulations is needed to achieve AI governance. 

In addition to the structure, an AI lifecycle is needed to achieve AI governance. The lifecycle of an AI system deployment consists of data collection, feature selection, model building using models chosen, testing and deployment stages. Governance at each stage of this lifecycle is needed to ensure that the system is within scope, cost, and risk. Furthermore, proper management of project’s team members is often ensured by policies that applies to each member.

Advantages of AI Governance

An advantage of AI governance is that all stages of an AI system deployment is visible to all members. With governance, the metadata of each stage is captured, and risks involved at each stage is visible to each member. This reduces risks involved in managing costs and time at latter phases as risks are visible from the first phase. The structure could improve its efficiency and reduce further risks when each phase is seen while minimizing schedule and cost overruns.

AI Governance also improves communication between team members. When policies are used in the AI system deployment each member has a set of tasks to perform under certain rules. This causes each member to contribute and collaborate with others during tasks they perform. When there is proper communication between members of teams, conflicts between them could be minimized and risks of project failure could be reduced.

AI Governance also reduces effort for higher level managers to monitor systems. Most AI systems used today use central document storage to store progress of deployments and maintenances. Most documentations are shared between different team members and each team should comply with the inputs of others based on policies of AI governance. This reduces the monitoring effort and effort of teams in requiring information to perform certain stages of a deployment as they rely on other teams.

Image Courtesy:

Comments are closed.