| | JULY 2021 9AI is not just increasingly prevalent, it's often invisible. It can be hidden in everydayHOW AI CAN REVAMP A BUSINESSbetween the technology department and the rest of the business. Similarly the Chief Risk Officer is the executive accountable enabling the efficient and effective governance of significant risks, which must include AI. This isn't only the AI developed by the business, but includes that integral to third-party services utilized by the business. Controlling how AI solutions are built, bought or rented, implemented and governed is typically the responsibility of the data scientists, data engineers, developers and the business owner. In a traditional risk management framework, there are often too few AI-specific controls.Few organizations have detailed actionable AI-specific ethics frameworks. AI audit criteria and ethics are often discussed but more rarely integrated, actioned, or enforced. The significant disconnect between those planning, building and deploying AI and those ultimately accountable for the risks in the eyes of regulators, customers and society is concerning.Ensuring Responsible AI, including integrating risk mitigation and attending to any ethical concerns, is essential for building trust in its application. It includes using explainable or auditable algorithms, embedding controls to catch undesired data and algorithm biases, considering the impact on privacy and supporting global standards of practice. Responsible AI, and in fact the responsible use of advanced analytics and data, must consider the complete ecosystem--from collecting and sourcing data, to the development and deployment of algorithms, to the actions that result from their outputs.Our solutions for ensuring Responsible AI are underpinned by five pillars:1. Is AI unbiased? Is it fair? There are many types of bias, and fairness is difficult to define, measure and control. While it may be difficult, it is not impossible.2. Is AI interpretable? Is it transparent and explainable? Can it be proved that it does what it is supposed to do? The degree to which an AI must be interpretable will depend on its purpose and, increasingly, on legislation and customer and citizen expectations.3. Is AI robust and secure? Is it safe? Working with this aspect of Responsible AI will require rigorous validation, monitoring and maintenance.4. Is AI appropriately governed? Ensuring this will begin at design and influence its operating models and processes.5. Is AI legal, ethical and moral? These are challenging questions and will depend on jurisdiction, company roles and moving societal norms. They also require developing, deploying and governing effective and actionable ethical frameworks in regard to AI and other advanced uses of data.The take away from this article for the CIO is: · As the use of AI increases in breadth and speed to market, business should prioritize its governance in accordance with tailored ethical frameworks;· Establishing a governing body, best-practice playbooks and risk documentation will reduce risk and evaluation time; and· Good AI governance will allow companies to use AI efficiently and safely to realise their ambitions.
< Page 8 | Page 10 >