The disruptive power of AI already affects a variety of roles, including customer support, operations, brand management, people and community, and, more recently, risk management and compliance.
FREMONT, CA: Artificial Intelligence (AI) has become an essential element for companies across industries. AI is driving business value and, as a consequence, is rapidly being adopted all over the world. Last year, the McKinsey Global Survey reported “a closely 25 percent year-over-year increase in the use of AI in standard business processes”. The disruptive power of AI already affects a variety of roles, including customer support, operations, brand management, people and community, and, more recently, risk management and compliance.
The need for enterprise audit software for AI systems
Documenting the actions of all AI solutions utilized by an organization: This includes tracking AI solutions and evaluating their distribution of features to investigate statistical dependencies. Considering the case of an AI solution for hiring, one must have clear insights into which attributes have the strongest impact on the recommendations.
Analyzing compliance with a set of specified requirements: Once one acknowledges the outcome of the model, it is necessary to determine compliance with particular requirements that may vary from legislation to organizational guidance.
Allowing for cross-departmental collaboration: This audit software should facilitate multi-stakeholder collaboration – in particular between risk managers and data scientists overseeing AI solutions – by providing appropriate information. For example, risk managers need non-technical explanations as to which requirements are fulfilled or not, while data science teams can be more interested in the performance characteristics of the model. When an issue of non-compliance is identified, the audit software should make recommendations to the technical teams for appropriate interventions.
Developing such audit tools for AI systems will go a long way towards reacting to the risks associated with AI. However, the responsible AI cannot be completely automated. There is no standardized list of criteria that must be met to minimize both current and potential risks, as the context and the business domain can often dictate which things are required. As a result, risk managers and their ability to exercise judgment will remain key. The emergence of AI would only encourage them to concentrate on what they do best: collaborate with other colleagues across departments in the design and implementation of sound risk management policies.
