The secret to any successful business isn’t money or building the best product.
It’s trust.
This month’s edition of Voices of Trusted AI is centered around the importance of building trust between AI, your business, and your consumers.
Tell me: What steps does your organization take each day to cultivate lasting relationships? Do you have contingency plans in place to prevent that trust from being broken?
Thanks for reading, Cal Al-Dhubaib, CEO
What Do We Mean by Trusted AI?
Trusted AI is the discipline of designing AI-powered solutions that maximize the value of humans while being more fair, transparent, and privacy-preserving.
What it's about: This month, President Joe Biden unveiled a new AI Bill of Rights, outlining the five principles that should guide the design, use, and deployment of automated systems:
Safe and effective systems.
Algorithmic discrimination protections.
Data privacy.
Notice and explanation.
Human alternatives, consideration, and fallback.
Why it matters: Despite our vast number of powerful tech and AI companies, the U.S. is one of the only Western nations without clear guidelines and regulation on protecting citizens against the unintended consequences of AI.
The AI Bill of Rights is intended to function as a call to action for technology giants, data scientists, and any company deploying AI to build these protections for those impacted by AI. Although many critics don’t believe these guidelines are stringent enough, the guide sends an important message: It is now more important than ever to design AI with privacy, transparency, and trust at the center.
Similar to what we saw during the release of GDPR, companies that proactively address and mitigate these concerns throughout their AI design will be able to navigate future AI regulation much easier than their dismissive counterparts.
What it's about: McKinsey surveyed more than 1,300 business leaders and 3,000 consumers globally to better understand the importance of digital trust and its impact on companies’ bottom lines. The results? A growing number of consumers make decisions based on their level of trust in a company. When data use and privacy policies are not clearly communicated, a majority of consumers will take their business elsewhere.
Why it matters: Today, company success hinges on the level of trust cultivated with consumers. Companies must transparently share how consumer data is used and protected if they hope to succeed in today’s AI-powered climate.
So, why are most businesses failing to mitigate digital risk?
Less than a quarter of executives reported that their organizations are actively mitigating digital risks, such as those posed by AI models, data retention and quality, and lack of talent diversity. Moreover, 57% of executives reported that their organizations suffered at least one material data breach in the past three years.
There is a major disconnect between the level of protection consumers expect and the level of action companies are willing to take. As we’re seeing with new government guidelines, it’s only a matter of time before mitigating risk is required rather than recommended. Will your organization be prepared?
What it's about: IBM Sales Leader, Stephan Schnieber, led a trustworthy AI workshop with his colleagues at the renowned Data Natives Conference in Europe. Throughout the session, the team unearthed how trust is defined, top challenges of trustworthy AI, and how trust can be improved through four key areas: bias detection, data drift, transparency, and explainability.
Why it matters: While all of Schnieber’s points are important, his team highlights one that is often overlooked when strategizing and designing AI: model drift.
Unlike some technologies that companies can “set and forget,” AI must be managed and updated often. As environments change (think: the emergence of COVID-19) and consumer behavior shifts, so too must your model. To continue to accurately evaluate and predict patterns, models must be retrained and redeployed.
Whether you adopt IBM’s points of establishing trust or follow anotherTrusted AI framework, AI design must be guided by fairness, transparency, and the preservation of privacy. Without proactive actions in place, companies are risking not only their consumers’ trust, but their safety as well.
Podcast: Change Management Prepares Leaders for AI
What it's about: In this podcast, Pandata CEO and AI Strategist, Cal Al-Dhubaib, sits down with entrepreneur and author, Scott J. Allen, to discuss how leaders must use change management principles to prepare themselves for AI-powered disruption.
Why it matters: The vast majority (read: 87%) of AI projects never make it into production. And while a number of factors impact this low AI success rate, many can be attributed to poor change management.
Leaders that take the time to improve their team’s AI literacy, identify specific use cases for machine learning, and set realistic expectations for their AI project, will see greater success than those that do not.
Moreover, effective change management stems from transparency and vulnerability. Being transparent with your organization and stakeholders makes communicating project results, changes, and failures much easier.
Have you prepared your team for trustworthy artificial intelligence?
What's one piece of advice for a company looking to deploy AI?
Decompose the business problem into a specific application scope.
There have been too many times we've heard a senior leader discuss implementing a machine learning solution to a problem without realizing it is three or four separate smaller issues to address.
For example, an organization wants to improve a vehicle routing problem with machine learning and expresses their need for "better optimization."
However, when probing for more information on the type of optimization the organization was looking for, they may discover a need for both vehicle path optimization and entire transportation network optimization.
The organization and team of data scientists will recognize these problems as completely separate machine learning projects, and can scope the work appropriately.
