IBM has recently launched its inaugural IBM Women Leaders in AI, in recognition of women advancing their companies journey to AI across diverse industries around the globe. In this blog, learn more about one of these remarkable leaders who is steering responsible AI with IBM Watson OpenScale.

“Only about 5 percent of our clients are heavily adopting AI at this point, and it begs the question of why and what’s holding them back? And what we found is that it’s really all about trust.” — Kelly Combs, KPMG

AI is all about trust

Kelly Combs is a Director of Emerging Technology Risk at KPMG. Her expertise encompasses robotic process automation, artificial intelligence, and how to responsibly govern, scale, and manage technologies under the umbrella of intelligent automation.

Over her nearly 10 years at KPMG, Kelly has held a birds-eye-view over the advent of modern AI, and reasons there are some clear factors to why some companies embrace AI where others may shy away.

According to Kelly, one thing that holds back adoption is the AI skill barrier: “AI has a reputation as a black-box,” said Kelly. “In order to get stakeholders’ and business users’ buy in—and ultimately a system that your customers can trust—you need to be able to have explainability and transparency into how the system is working.”

Perhaps, in your office, you’ve seen Kelly’s words at work: companies stall or slow down their uptake of AI due to a lack of trust. Most AI adopters are in pursuit of certain assurances before moving forward with AI, including things like security by design, or in the case of financial institutions, injecting traditional models with AI algorithms that are unveiled and explainable to better review or measure fairness, accuracy, and performance.

Kelly named four trust imperatives that, when addressed, will accelerate AI adoption and spur companies to begin operationalizing AI in their business.

• Integrity: How do we ensure data quality throughout its lifecycle?
• Fairness: How do we reject prejudice or bias towards groups, sets of individuals, or data attributes?
• Explainability: How can we in business terms explain the decisioning of how AI came to its conclusion?
• Resiliency: How can we shield AI and AI infused services against cyber threats or adversarial attacks?

KPMG earns client trust with Watson OpenScale

Most recently, Kelly and her team at KPMG focused on how to solve their clients’ unique governance needs and satisfy these four trust imperatives. Kelly concludes that the days of manual governance are coming to an end: “We need to start to digitize the way that we monitor and evaluate AI models in real-time with tools that give us insight versus doing point-in-time, manual assessments,” said Kelly.

Take determining the credit worthiness of a customer. Current AI models use data attributes such as FICO score, income level, or age to help make predictions to approve or deny a loan or line of credit. But according to Kelly, evaluating AI manually is not enough to keep tabs on when models are beginning to drift, or start favoring certain outcomes or individuals, thereby breaking the concept of fairness.

To solve this, KPMG turned to Watson OpenScale, which Kelly called a major differentiator in the AI marketplace.

“Watson OpenScale is one of the only technologies in the marketplace that gives transparency in business terms to our clients by articulating how and why an AI model came to its determination as well as what data attributes were used.” — Kelly Combs, KPMG

Watson OpenScale helps build and scale transparent and explainable AI in the enterprise and provides insight into which data attributes are contributing to the decision-making of an AI model. This level of transparency has proven extremely valuable to KPMG and is used to help clients better understand unconscious bias, be more proactive to regulatory change, and provide the trust they need to drive forward AI adoption. See why Watson OpenScale won a Firestarter award for innovation from 451 Research with this report.

Accelerate your journey to AI with a prescriptive approach. Visit ibm.com/data-ai to learn about how IBM’s ladder to AI helps you modernize, collect, organize, analyze, and infuse all your data.