On a Monday morning at 6:30 AM, Josh sipped his latte, his elbows atop the service counter. Each day at this time, the sunlight through the front window blanketed his coffee shop and he enjoyed a few moments of peace and quiet before the morning rush began. With 30 minutes to spare before he corralled his baristas for their morning pep talk (and shot of espresso), he unlocked his iPad and pulled up his most valuable assistant: his small business dashboard. In seconds, Josh’s supply advisor scoured his accounts, sales and expense histories, local weather forecasts, event information, and past tourism data, and told him he would need five new sets of filters and 1,000 plastic cups for the coming week. He ordered them from Amazon with a single touch.
Josh also knew the shop needed a new espresso machine, but he had been putting it off for over a month. With the savings in his account, he could either order the new machine now or make a payment on the term loan he had taken out two years ago to start the business. If he continued to put off a replacement, the machine could break at any moment, and espresso was the second-best selling item on the menu (after iced coffee). On the other hand, he was almost done paying off his loan, and procrastinating another month would add interest.
Josh asked his robo-adviser for advice. “You can do both,” it reported. “Given your expected sales for the month, it looks like you’ll be able to use your savings to pay down the loan and put the espresso machine on your credit card, which has available credit of $3,500. When the credit card payment comes due in 30 days, you will have the cash to pay it off, based on current sales projections.” With one tap, Josh purchased the espresso machine. After the morning pep talk with his staff, he opened the doors for the day, confident in where his small business was headed.
At the end of the day, as Josh was closing up, his bot reminded him that it was June 1, and that quarterly taxes would soon be due. He momentarily worried that he had overlooked his tax payments when buying the new espresso machine, but then the bot said, “Don’t worry. Your estimated tax payments have already been accounted for in your cash projections for June.” Finally, with a few more taps and swipes, Friday’s payroll was set, healthcare deductions were taken from his employees’ paychecks, and taxes were ready to file.
Josh’s story is fictional, but it points toward the potential for technology and specifically artificial intelligence to assist entrepreneurs and improve the small businesses they run. Make no mistake: these technologies are already changing business, and so far they seem to be providing an advantage to large companies. The question is whether they can be designed and adopted to help small businesses, not just giant companies – and on that front I am hopeful.
As technology opens the doors to vast troves of data, opportunities are emerging to create new insights on a small business’s health and prospects. Insights from this data have the potential to resolve two defining issues that have faced lenders and borrowers in the sector: heterogeneity—the fact that all small businesses are different, making it difficult to extrapolate from one example to the next—and information opacity, the fact that it is hard to know what is really going on inside a small business.
From a lender’s point of view, the smaller the business, the more difficult it is to know if the business is actually profitable and what its prospects might be. Many small business owners do not have a great sense of their cash flow, the sales they might make, when customers will pay, or what cash needs they could have based on the season or a new contract. Small businesses have low cash buffers, so a miscalculation, a late payment, or even fast growth could cause a life-threatening cash crunch.
But what if technology had the power to make a small business owner significantly wiser about their cash flow, and a lender wiser as well? What if new loan products and services made it easier to quickly and accurately predict the creditworthiness of a small business, much like a consumer’s personal credit score helps banks predict creditworthiness for personal loans, credit cards, and mortgages? What if a small business owner had a dashboard of their business activities, including cash projections and insights on sales and cost trends that helped them weave an end-to-end picture of their business’s financial health? What if this dashboard helped them understand all credit options they qualified for today and which actions they could take to improve their credit rating over time? And better yet, what if the dashboard, marshaling the predictive power of machine learning amassed from data on thousands of business owners in similar industries, could help a business owner head-off perilous trends or dangers?
This future is appealing because it responds to the fundamental need of small business owners to be able to see and more clearly interpret the information that already exists, helping them navigate the uncertain world of their businesses on their own terms and plan accordingly. And it provides an opportunity for lenders to better understand the creditworthiness of their potential customers and lower lending costs as a result. I call this imagined future state “Small Business Utopia.”
At the same time, it is easy to imagine a dark side to the advances in technology, particularly artificial intelligence. Economists have begun to explore the implications of artificial intelligence on innovation. They view it as a “general purpose technology,” which has the potential to create significant advances in multiple industries. They also suggest that the winners are going to be those who have control over large amounts of structured and unstructured data.
This raises a potential risk of artificial intelligence. If certain companies are allowed to have a monopoly over collections of data, this could adversely affect future innovation and the shared benefits it would bring. Future regulation needs to ensure that there is open access to data streams to power better insights for small businesses and other sectors. For example, the UK has implemented Open Banking, a policy that clarifies that customers, including small business owners, own their banking data and allows them to grant third parties access to this data to create innovative new products and services.
In addition, as machines learn to identify who is more likely to default on their loans, the risk of discrimination and exclusion becomes significant. Most worrisome is the idea that these decisions will be made by a “black box”; no one would know exactly which data attributes the machine was using to make recommendations or decisions. A machine might identify a risk factor that happens to correlate strongly with race, gender, or the characteristics of other protected classes and—barring explicit rules preventing it from doing so—include it as a pricing factor. More generally, sophisticated algorithms that perform exceptionally well along some narrow dimensions yet lack intuition and situational awareness could create serious problems.
Black box models may be incomprehensible, but that doesn’t mean they can’t be audited. Both companies and regulators will need to develop new methods to untangle the inner workings of the algorithms of the future. Even if automation is developed that is capable of detecting discrimination and other bad outcomes, it seems likely that human oversight will still be required – both within companies and by regulators.
The use of big data and algorithms will bring new products and services, but also bring some new concerns. It is not yet clear what impact the changes we anticipate from technology will have on access to capital for traditionally underserved markets. In the past, women and minorities have struggled to find willing lenders. The hope is that with more efficient markets and new data sources, more creditworthy borrowers from underserved segments of the market will get loans. However, “black box” algorithms, where the formulas are not open to review, could lead to more discrimination, not less.
One way to get ahead of these concerns is to collect the actual data on access to capital in the small business market. The most relevant metrics would be loan origination data by size of loan and by type of small business owner. Section 1071 of the Dodd-Frank Act—the law requiring this data collection—was passed after the financial crisis but has yet to be implemented. More innovation can take place if there is a way to track market outcomes. Collecting this information and using it to identify and correct market gaps is a critical foundational element of a highly functioning small business credit market, and it will only become more essential as artificial intelligence becomes an integral part of lending decisions.