Can The Board Truly Govern AI? - Mark A. Pfister
The Growing Dilemma of the Board’s Grasp of Artificial Intelligence Risk
(Originally appeared in the November 17th, 2021 ‘Across the Board’ digital publication, a Board Director, Board Advisor, C-Level, and Business Leader publication reaching 27,500+ exceptional business leaders in over 70 countries with articles focused on leadership, strategy, and governance topics — sign up here)
It is almost impossible to get through even a single day without hearing about Artificial Intelligence or its truncated acronym, AI. What started as a movement of disparate, yet useful algorithms has quickly evolved into a complicated and interrelated web of 0s and 1s trying to solve the world’s biggest challenges on a scale unfathomable just a few years ago. With the ability of AI along with Machine Learning (ML) to allow machines to learn from their own experiences, the capacity to adapt to new and continually changing environments increases exponentially over time. By assembling and analyzing large amounts of data, this allows AI/ML to identify patterns and react with growing accuracy and effectiveness.
Essential Meaning of Artificial Intelligence (AI):
1: an area of computer science that deals with giving machines the ability to seem like they have human intelligence.
2: the power of a machine to copy intelligent human behavior.
Note the phrase ‘seem like’ in the first definition. This is very different than stating that machines ‘have’ human intelligence. We should be paying very close attention to this nuance, especially as it relates to the governance approaches and oversight duties of the Board — the purpose of this article.
Beyond the responsibility of programmers to understand the use, ethics, and safety mechanisms in their AI/ML creations, Boards of Directors also have a lofty responsibility in the decision to rollout, oversee, and govern AI/ML in their organizations. Tangible and transparent governance models which operate efficiently in a pre-AI/ML environment can quickly become opaque and secretive in a post-AI/ML environment, camouflaging important areas for a Board and the overall organization to monitor & assess.
Boards are already starting to struggle with this monumental governance task, not solely due to the inherent complexities, but also due to a legacy weariness to adapt to “yet another change” from the previously steady-state, unchanging boardroom of yesteryear.
To be clear, I am not down or negative on AI or Machine Learning, in fact it is quite the opposite. AI/ML usefulness and opportunities are endless along with the ability of this technology to ‘do good’ on a global scale. However, Boards of Directors have many moving parts to consider to fully live up to their duty of care responsibilities.
Outside of the ubiquitous and oft discussed ethical, data privacy, and workforce-eliminating concerns of AI/ML, we are simultaneously witnessing a governance revolution on a global scale. For many Boards, this revolution is silent until it is not — meaning that it can become an all or nothing event …and by the time it is an ‘all’ event, it is already too late. Although there are numerous examples of AI/ML risks becoming reality in multiple geographies, a high-profile AI/ML event was recently witnessed in the U.S. with catastrophic fallout. This AI/ML misfortune happened at Zillow, the well-known real estate listing company.
Zillow touted such a deep confidence early in 2021 in its ability to accurately estimate the value of homes that it offered a new option within its existing Zestimate estimating product called Zillow Offers. For certain homes, this new option made home sellers eligible to receive a direct cash offer from Zillow to purchase the property. Interesting and innovative concept, indeed, and likely a major competitive industry differentiator. With the primary intent of streamlining the home-selling process for sellers, Zillow infused itself right into the home buying process as an actual home buyer as part of its home-flipping line of business. The timing of this offering for home sellers seemed perfect, especially with the ability to potentially eliminate face-to-face interaction throughout the pandemic. Skipping right to the end of the story, just eight months after its launch Zillow is shutting down the Zillow Offers division entirely while trying to unload over 7,000 properties at a discount in an effort to recoup a portion of its losses.
The decision to shutter Zillow Offers is a stunning announcement for the company and a wakeup call to all industries, not just the real estate industry or ‘iBuyer’ industry vertical. Following a $304 million inventory write-down in the third quarter of 2021, Zillow also announced its plans to eliminate 2,000 jobs (roughly 25% of its staff) while simultaneously experiencing continuing stock losses.
…But what exactly created this deep sense of security within Zillow to believe that it could confidently and quickly offer not only top dollar for home purchases, but also pay above-market offers to secure the deals? Well, that’s exactly where the AI/ML component comes into play…
Sure, the AI/ML in Zillow’s example was designed to assess a vast amount of information very quickly when calculating a fair offer for a home, and weighed factors such as comparable home sales in the area, timeframe of local home sales, desirability of specific neighborhoods, among other countless traditional inputs. However, what can’t be accounted for specifically (yet) in this AI-leveraged scenario includes the less subtle indicators that a seasoned real estate expert would recognize based on years of experience. These nuances can have drastic impacts on the overall house value, not to mention the amount of capital investment required to get the house to a desirable or sellable state. Additionally, market fluctuations and areas such as consumer/purchaser sentiment (the ‘feel’) many times can’t be forecasted.
The Zillow example is just one of many confidence and value-erosion instances we will witness in upcoming months and years as it relates to improperly applied or governed AI/ML. I have personally pointed out many of these growing AI/ML governance blind spots for numerous Boards in recent months as they are sometimes much easier to identify when looking in from the outside.
Without oversimplifying the issue, there are indeed some simple steps for a Board and Board Member to not be blindsided by AI/Ml hocus pocus. In an effort to lessen or eliminate the technical fear (which, by the way, is the number one Board hindrance) that commonly disassociates the AI/ML area from the Board’s obvious governance focus, the following list can greatly help:
- Has the Board erroneously delegated the full AI/ML decision-making or governance to the technology vertical within the organization or Board?
- Regardless of the data inputs, do the outputs of the AI/ML decisioning components pass the “layperson’s smell test”? If a suggested direction or output seems off or illogical, it likely is… Dig into it.
- Don’t be fearful to ask any questions relating to AI/ML, even if you think they may seem obvious to other Board Members. Other Board Members may simply feel too exposed or embarrassed to ask.
- Tie the data & decisions that the AI/ML spits out in your organization to a dashboard that directly compares these decisions to industry benchmarks and trending. Sluggish performance or lagging indicators may indicate hidden risks. On the other hand, if your ‘competitive edge’ or performance seems too good to be true, evaluate if this is due to short-term anomalies (potentially hiding larger long-term issues) or true innovation.
- Understand that AI/ML is currently being applied both within the boardroom as well as within the organization — separately. Educate yourself on the difference and the nuances of both.
- Don’t assume anything!
The leveraging of AI/ML to enhance decision-making inside and outside the boardroom is expected to accelerate exponentially in upcoming years. As it relates to proper governance, Boards must adapt quickly to the growing AI/ML expectations from shareholders (competitive advantage) and stakeholders (innovation to combat global challenges). Boards that embrace this direction and start as early as possible will likely find themselves ahead in their industry while also lowering their organizational risk.
Is your Board prepared to effectively govern AI & ML risk?
Reach out if you need help in building or validating your Board’s governance model with our Consulting & Advisory offerings, Board Director Education & Certification program, and International Speaking Tour topics.
Non-Executive Director | CEO | Chief Board Consultant | Corporate Strategist | Board Macro-Influencer | Speaker | Author — www.PfisterStrategy.com
About the Author: In addition to serving on numerous Boards, Mark A. Pfister is a renowned Board Consultant, ‘Board Macro-Influencer,’ certified Board Director, speaker, author, and advises public, private, and nonprofit Boards in efficient and effective operations. Known as ‘The Board Architect,’ he is also the inventor of the ‘Board as a Service’ (BaaS) engagement model and an expert Project Executive frequently advising on strategic global initiatives in their initiation and operational phases…… << read full bio here >>
Mark A. Pfister - The Board Architect
Non-Executive Director | CEO | Chief Board Consultant | Corporate Strategist | Board Macro-Influencer | Speaker |…