AI Governance

To meet the need to provide higher quality and richer services in a cost-efficient manner, Communications Service Providers (CSPs) must adopt complex new technologies while simultaneously reducing costs. To do this, the industry relies on AI to effectively augment and automate decision-making that would overwhelm human operators. It is estimated that by 2025 there will be 30+ billion connections worldwide. Manual means of assurance cannot scale to satisfy such digital demands.

Keeping track

Given that CSP Business and Operations Support Services (BSS and OSS) architectures typically contain hundreds or thousands of significant components, it is probable that an equal number of AI modules will eventually be deployed within a CSP organization. Managing AI at this scale leads to accountability, audit and maintenance problems. For example, if it is discovered that a data set used to train AI is corrupted, it is natural to withdraw and redevelop all the models affected - but which ones are they, and where are they located? Can the CSP demonstrate to regulators that every model involved has been removed? Can these tasks be performed rapidly and with little cost?

Standards and best practices are needed

AI is much like any other business intervention: organizations want to know that it is:

• Effective – it does the job it is intended to do
• Safe – it is predictable and can be controlled
• Proportionate – it achieves its role without undue effort or costs

Organizations can ensure that these objectives are met by adopting AI Management Standards. Such standards help provide a framework that demonstrates proper control of AI to the satisfaction of internal stakeholders, external regulators and customers.

Moreover, it should be practical to apply AI Management standards across the full range of use cases that the AI is expected to address – from chat bots to 5G network optimization.