Article from Harvard Business Review written by Douglas Merrill on MAY 24, 2019. Being a board member is a hard job — ask anyone who has ever been one. Company directors have to understand the nature of the business, review documents, engage in meaningful conversation with CEOs, and give feedback while still maintaining positive relationships with management. These are all hard things to balance. But, normally, boards don’t have to get involved with individual operational projects, especially technical ones. In fact, a majority of boards have very few members who are comfortable with advanced technology, and this generally has little impact on the company.
This is about to change, thanks to machine learning and artificial intelligence.
More than half of technology executives in the 2019 Gartner CIO Survey say they intend to employ AI before the end of 2020, up from 14% today. If you’re moving too slowly, a competitor could use AI to put you out of business. But if you move too quickly, you risk taking an approach the company doesn’t truly know how to manage. In a recent report by NewVantage Partners, 75% of companies cited fear of disruption from data-driven digital competitors as the top reason they’re investing.
The questions boards are going to have to ask themselves are similar to those they would ask in the face of any large opportunity investment: Why are we spending all this money? What’s the economic benefit? How does it impact our people and our long-term competitiveness?
Answering these questions requires expertise in technology. But you can’t just add a tech expert to the board and count on him or her to keep the rest of the board up to speed. Having served in that role, I have found it to be at best a useful half-step. Relying on a single techie is no replacement for having a full board mastering at least a basic understanding of AI and its disruptive potential.
Every board’s comfort level is going to differ depending on the industry. Manufacturers well understand how robots can free up people to do higher-order work by taking on repetitive and potentially dangerous jobs. Hospitals and health insurers are starting to deploy AI widely, but big successes have been elusive. By contrast, the financial services business is ripe for disruption by AI. Lenders have massive amounts of data and the potential to free up billions in cash flow by finding new efficiencies through applications that will, for example, help bankers make smarter lending decisions and create new revenue opportunities by offering customers better, more tailored products.
That said, here are four guideposts that board members in any industry can use to orient themselves when they begin the journey:
It’s math, not magic. Boards shouldn’t be intimidated by AI. Members don’t need to have degrees in computer engineering to understand the technology behind AI, just like they don’t need to be CPAs to understand the company’s balance sheet. Any good use of ML or AI is going to be an outgrowth of what the company is already doing, not some kind of universal all-knowing Skynet type of AI. Keeping that perspective at the forefront and gaining a basic understanding of AI will help boards better decide how to direct AI use.
Well-run AI projects should be easily understood. When evaluating if a project is right for their company, boards should feel confident enough to say when something doesn’t make sense. The best-run AI projects should be explainable in plain English. It should be clear how real groups of people, whether employees, customers or management, will be affected. If a vendor or internal team can’t explain how an AI project works, it may not be the right fit for your company. This is not unique to ML — it used to be true for ERP implementations — but ML is moving more quickly through the corporate world than ERPs did. For example, when I presented an ML-underwriting project to the board of one top credit-card issuer, I started with the economic impact to their business, the timeframe for delivery, what the roadblocks might be for IT and compliance, and who would need to get involved.
You don’t have to get creepy to get value out of data. Too often, companies assume that in order to make the most out of AI, they need to be like Facebook or Google and pull in every last bit of data they can find. But that can get creepy fast and, usually, there’s no need for that level of data. Our work developing machine learning-based credit underwriting models with banks and lenders has shown that social media data doesn’t provide such strong signals, anyway. Most companies are already sitting on a ton of pretty banal data that’s full of signal and insights that can be unlocked using ML.
AI is an operating expense, not a capital investment. If management’s plan for getting on the AI bandwagon revolves around a big one-time investment, chances are they are going about it wrong. AI has the potential to enhance the bottom line by boosting revenue and cutting costs, but budget needs to be put aside to ensure the algorithms and models are functioning properly and are being rebuilt or refit as macro factors change and new sources of data emerge. Think of AI as you would a Formula 1 race car, which performs best when its support team has a real-time view of the vehicle’s health as it’s zipping around the track.
Widespread adoption of AI in business is still in its infancy. Boards that fail to get in front of this trend will pay the price.
Douglas Merrill
Douglas Merrill is the CEO and founder of ZestFinance, a Los Angeles-based financial services technology company. He was previously CIO and VP of Engineering at Google.
