This fall, Basis Technology is sharing a series of blog posts exploring AI and the “black box” problem: the regulatory and implementation barriers caused by the un-explainability of sophisticated AI.
In this week’s post, Peachtree AI’s Joe Sutherland explores AI’s core value proposition in highly regulated industries.
The Essential AI Value Prop: Lower Costs & Higher Revenue
There’s a moment that everyone has when they’re in the middle of a technological revolution and they realize it. You can imagine the first Model T’s rolling out of factories and onto streets having this kind of unique effect on onlookers as an idea took shape: Horses, once the foundation of personal transport, would never truly be needed again.
Not that long ago, I had a similar realization. When I was in graduate school, I worked for a startup that transformed central bank communications into tradable information. To do this, we built an AI system that analyzed the historical relationship between the Federal Reserve’s or the Bank of England’s language and market pricing. The AI then used that analysis to forecast how future communications would impact markets. Sometime during this process, I was struck by the larger implications of our work: AI could now do higher level processes—something that once only a Ph.D. could do—only cheaper, faster, and (to be honest) with less ego.
The point here isn’t that AI will take over white collar work. That would require another order of technology—artificial general intelligence (AGI)—and we’re not even close to that. The point is that AI has arrived, and it presents a powerful value proposition we’re only beginning to understand.
That said, the broad strokes of its impact are known. AI excels at prediction, and, as the technology is adopted by industry, it’s going to profoundly reduce the costs entailed in running a business while creating opportunities that significantly boost revenue.
Prediction: The Core of the AI Value Proposition
Prediction is fundamental to human activity. Every choice we make requires that we make a prediction about what the outcome of that choice will be. For example, life comes with quite a few big decisions: Whom should I marry? Where should I go to school? Which house should I buy? When making these big decisions, we first predict the potential outcomes, and then we try to make the choice that brings us to the best outcome.
We run similar tests even for the smaller choices we make. The answers to questions like, “What movie should I see tonight?” or “When should I go to lunch today?” are only arrived at through a comparison of forecasted costs and benefits. Every time we engage with even the tiniest of decisions, we pay a cost in the form of time, anxiety, and the effort expended to accurately run the prediction. Those costs may seem trivial, but they add up. A recent TIAA study demonstrated that Americans actually spend more time on little decisions like choosing a restaurant than they do on major decisions, like planning their IRA investments or buying a house! Decision fatigue from the many thankless tasks with which we are faced every day – substituting semicolons for commas, logging prospect information into CRM, producing memos summarizing monthly work – can cumulatively overpower our ability to make the most important decisions of our lives.
The power of AI is to make cheaper, faster, and better predictions that supplant our efforts to evaluate alternatives and make judgments. By scaling choice evaluation and automating smaller decisions, we will make better choices on the things that matter. For businesses, this labor saver will profoundly affect the essential components of profit: costs and revenue.
Because access to faster, better, and cheaper predictions will raise efficiency, the most significant contribution that AI is making in industry is the reduction of costs. Should this document go to John or to Jane? How should I replace this missing data in the spreadsheet so that I can run my sales predictions? Does this widget look broken or not? Are supply chain issues about to crop up in this region? AI can and will continue to make more mini-decisions like these easier…or take them over altogether.
I remember consulting for a large company that needed to sort documents into several categories. For this simple classification task, the firm had a dedicated staff of six full- time employees. Using the available information, I trained an AI application to sort tens of thousands of documents more quickly than the entire team combined—while maintaining and exceeding the quality of their work. The cost of running the system was also less than one-sixtieth the cost of the team, producing significant business savings and freeing up those employees to do more effective, human-level work for the business.
For firms in highly regulated industries, this aspect of the value proposition is particularly attractive. While they often protect consumers, regulations regularly translate to millions (sometimes billions) in lost profits for businesses through paperwork, personnel, and fines. Understandably, companies in these spaces are keen to adopt techniques and technologies that allow them to mitigate these costs through improved efficiency. Consequently, it’s the unsexy tasks AI can automate—the “low hanging fruit”—that will likely move the needle most in finance, healthcare, and other heavily regulated industries.
About This Excerpt
The only way AI's going to make a real impact in finance, healthcare, and other highly regulated industries is if the "black box" problem tackled head on.
The Amazing, Anti-Jargon, Insight-Filled, and Totally Free Handbook to Integrating AI in Highly Regulated Industries does exactly that. Featuring in-depth pieces from almost a dozen subject-matter experts, this handbook provides a comprehensive breakdown of the problem… and detailed strategies to help you create a solution.
Improving Revenue by Seizing Opportunities
Although commonly overlooked, AI has significant potential when it comes to increasing the top line. Better predictions underpin fundamental strategic choices, and better choices mean bigger revenue.
Herbert Simon, the winner of the Nobel Prize in Economics, argued that humans are prone to satisfice (satisfy+suffice). Humans satisfice when they find the option that seems to reflect the best possible outcome over the search that was conducted. However, they might be missing out on a much better outcome simply because they were unable to search for it. Satisficing is a product of bounded rationality: the idea that humans can only consider so many options in the time given to them. The power of artificial intelligence in making strategic, long-lasting decisions is that we can analyze huge amounts of data and understand what the outcomes might be for all options before the choice is made. The best of the options evaluated might be good enough—but we should pick the best option overall.
AI products that “boil the ocean” of regulatory filings, for example, will help executives make better decisions because AI will uncover regulatory trends that they otherwise would not be able to see. In other words, I might not want an algorithm to decide whom I marry, but it would be very useful for it to suggest people I should think about dating. We wouldn’t want to miss out on the people we could have the happiest life with just because we couldn’t find them.
There is a flip side, however, to assessing every possible alternative and choosing favorites. The amount of data AI can analyze is so enormous that it’s not uncommon for the machine to report a significant relationship between two variables or factors when in reality such a relationship is entirely spurious.
An AI system could analyze Atlanta arrest records and observe a correlation, for example, between ice cream and crime. This output could encourage a less-than-careful analyst to infer that increasing ice cream consumption causes an increase in crime, missing out on the fact that it’s actually summer that’s causing both: People eat more ice cream during the summer, and crime is more prevalent during the summer because people are out and looking for things to do.
In other words, scientific rigor is integral when using AI: Correlations don’t necessarily mean causation. Given enough data, it’s easy to find false connections and make bad decisions.
Robots Mean You No Harm
Artificial intelligence isn’t a panacea, but it is causing a revolution. It’s doing this in two basic ways: by automating simple decisions and by providing valuable context for big bets. A future built on AI that works hand-in-hand with humans looks bright—but it looks brightest for the mavens who adopt AI first.
Early adopters will build significant and enduring competitive advantages by scaling slow and expensive processes with replicable and inexpensive AI. They will also leverage opportunities in positioning by capitalizing on market needs, monetizing untapped data streams, and capturing market segments that value artificial intelligence. While their peers are caught rewiring their models to adopt these new technologies, the mavens will already have edged out the competition and expanded into new markets.
Given its profound power to influence costs and revenue, artificial intelligence isn’t a project to be left wholesale to data scientists and IT. It’s a strategic business opportunity that requires technical know-how and organizational coordination. Companies making the transition to AI must build data and analytics into their core functions, an enterprise that requires careful change management and cross-cutting transparency.
Artificial general intelligence isn’t around the corner, waiting to form Skynet and take over the world. For the foreseeable future, the only thing a dreaded robot invasion portends is less meaningless work and more informed decision making. This is the AI value proposition, and I think it’s a compelling sell.
About the Author
Joe is an expert in leading groundbreaking technological initiatives, by building or supporting a data science team to deliver lasting results. A career in technical and operational roles at venues including The White House, VC-backed fintech ($20m), Columbia, and Princeton help him to represent both the front and back offices. He has extensive experience deploying production-grade machine learning architectures on cloud-based stacks (AWS/Azure/GCC) with various database infrastructures (SQL, Hadoop, Spark, NoSQL). His academic research, published in top peer-reviewed outlets, leverages predictive modeling (in R & Python) on structured and unstructured data to make behavioral inferences. He codes fluently in more than 8 programming languages and manages open-source software ranked higher than 99.9% of projects on GitHub. Ph.D., MPhil, and MA at Columbia and a BA at Washington University in St. Louis.
About Peachtree AI
Peachtree AI helps transform businesses into state-of-the-art, scalable organizations that profoundly outcompete on cost, innovation, and customer engagement. We work with clients who are technology mavens, with the potential and ambition to lead their industries, and with the adaptability to define the future. Our team, which has been helping clients globally with their technological needs for more than a decade, delivers results by leveraging significant experience in project management, analytics, and software development. With Peachtree AI, you can build a leading organization with a long-lasting competitive advantage.