The good news about AI’s risks? Companies are aware of them. The bad news? Most are not actually mitigating them. When we asked our survey respondents for their top-three priorities for AI applications in 2021, the top choice (picked by 50%) was responsible AI tools to improve privacy, explainability, bias detection and governance. But when it comes to action, only about a third report plans to make AI more explainable, improve its governance, reduce its bias, monitor its model performance, ensure its compliance with privacy regulations, develop and report on AI controls and improve its defenses against cyber threats. And in the case of explainability, companies have taken a step back compared to our 2020 survey.
Responsible AI is the only way to mitigate AI risks. When you use AI to support business-critical decisions based on sensitive data, you need to be sure that you understand what AI is doing, and why. Is it making accurate, bias-aware decisions? Is it violating anyone’s privacy? Can you govern and monitor this super-powerful technology?
AI’s data, technology and talent tend to be highly distributed across different functions and multiple third parties. You have to keep an eye on AI (and its data) from the beginning of model design through development, deployment and ongoing adjustments—because AI keeps learning and changing itself. Adding to the challenge: AI is a complex technology that many executives, including risk officers and even IT experts, don’t yet fully understand.
If your company is using AI, you need to make it responsible—right now.
Assess your risks and establish a plan to test and monitor.
Take a close look at how AI affects your financial, operational and reputational risks wherever you (or your partners) are using it. Update controls around its use accordingly, making sure they cover every stage of the AI life cycle—to support trust in your AI program.
Since AI keeps learning and changing itself, your governance has to function at AI speed. Your responsible AI toolkit must be always-on, always monitoring model performance, potential for bias and new sources of risk—and always adapting.
Operationalize your ethics.
Your AI has to represent your values, or it could (automatically) betray them. Create frameworks and toolkits to continually assess current and planned AI models, making sure they are not only explainable and robust, but also fair and ethical.
PwC’s annual AI Predictions survey, now in its fourth edition, explores the activities and attitudes of US business and technology executives who are involved in their organization’s AI strategies. Among this year’s 1,032 survey respondents, 71% hold C-suite titles and 25% were from companies with revenues of $5 billion and up. They are from the following industries: industrial products (20%), consumer markets (20%), financial services (18%), tech, media and telecommunications (17%), health industries (17%), and energy, utilities and mining (8%). The survey was conducted by PwC Research, PwC’s global Center of Excellence for market research and insight, in October 2020.