What Would Adam Smith Say About Your AI?

You remember him from high school economics. The wee Scottish lad that lived with his mother and changed the way the world looks at the economy. He threw together a five volume magnum opus titled An Inquiry into the Nature and Causes of the Wealth of Nations in 1776 while some other important document was being written across the ocean. Better known by its short title, The Wealth of Nations, Adam Smith created economic theories and frameworks that have stood hundreds of years of testing and new application. In 2025, Adam Smith’s views on economics still apply at the macro level, but they also have something to teach us about implementing AI. Many organizations across industries struggle with how to measure the expected value of a projected AI implementation into their workflows. But Adam Smith gives us some clues. Cozy up with a plate of haggis and your favorite peaty Scotch as we talk about Division of Labor for AI. Getting maximum returns from an AI investment was a perplexing question in 2024, but as we open the door on 2025, organizations are looking more critically at the enterprise AI market. Factors such as data privacy, on-premise deployments, and hallucination reduction are now front and center. These requirements for safe, secure, and trustworthy AI deployments create a central question: Should your AI be a generalist or a specialist? There are several AI tools on the market that are trained broadly with trillions of parameters and were designed to be as broadly applicable as possible. If you want a funny picture of Adam Smith, John Maynard Keynes, Milton Friedman, and other famous economists posing for a selfie? Coming right up! Need an outline for a screenplay that explores the juxtaposition of nihilism and hope through a love story between two dogs during the COVID-19 pandemic? Just wait a few seconds. Actually, that sounds amazing. Anyone out there work for Netflix? The point is that the AI models to which we are the most accustomed are built and designed to be general purpose and as broadly applicable as possible. They are also not designed to protect privacy nor with security in mind. These systems are generalists and, according to Adam Smith, will not generate the “wealth” that organizations are looking for. More on that in a moment. Creating a specialist AI not only answers major concerns such as privacy but also results in more valuable output, more transparency, and risk reduction. If Adam Smith could get a PhD in Applied Machine Learning and evaluate how we deploy our AI tools today, we could predict what he would say. The first three chapters of The Wealth of Nations specifically asks and addresses where wealth comes from. That’s the same problem leaders are trying to solve today whether in pure dollars or in mission value. But Adam Smith would say that “wealth” is stuff, not money. In this case, wealth from AI is the quality of the product that you receive as output from your AI deployment. So, the problem we need to solve in AI deployment is: How do we maximize the quality of the output of our AI relative to our organization’s mission? Back to our Scottish friend. Smith’s answer would be Division of Labor . According to Smith, Division of Labor is how wealth is created because laborers who specialize create higher quality products than generalists. Those higher quality products can be combined with other high-quality products to demand higher prices and thus more economic growth. This is increasing returns . Again, isn’t this what we are looking for in our organizations? An AI deployment is the step beyond Division of Labor in the Smithean sense representing the next level of value and productivity that Smith could have never envisioned. For those who deploy it correctly, it also means higher returns than Smith would have predicted. Share Human Labor, Specialized AI Highly profitable and complex organizations around the world hire specialists to execute their missions from analysts to mathematicians to marketing experts. The skills of those specialists bring unique economic and mission value, or the “stuff” that comprises wealth according to Smith. Organizations throughout the public and private sectors struggle with hiring the talent they need because they are looking for human specialists to fill critical and highly specialized roles. Putting significant effort into hiring specialist humans that are paired with generalist AI is not a recipe for success. Spending all the time, money, and effort to hire specialized critical thinkers should be paired with AI that will augment them and create “wealth.” The recipe for success is to find specialist AI that will augment human specialists and create more value per unit of work than a competitor. This is accomplished by implementing customizable, secure, and privacy preserving AI that is trained on the specific data and use case vertical that the organization needs. An AI system trained for use by criminal law attorneys should not be used by national security professionals any more than an AI that writes screenplays should be used. Quantitative hedge funds are not going to hire poets as portfolio managers so why would you do the equivalent in your organization? Read the post below if you’re just starting to ask what AI your organization needs. Thinking about AI implementation in the generalist versus specialist sense is informative. It helps guide organizations toward the right system for them, but it goes further. This approach also leads to better governance, more transparency, trust building, and the ability to measure your AI’s performance more accurately. While generalist AI led to innovations and the democratization of AI capabilities, it also has risks to implementation that create barrier to entry for organizations with sensitive data or missions. Instead, we should look to the Scottish Highlands and to our high school economics classes. Do you need a specialist or a generalist? If you need a specialist, don’t implement generalist AI. If you need privacy and security, don’t implement a system that was not designed for that. Create more value per unit of work by combining your human specialists with specialist AI and measure the output accurately. Specialized AI is the answer for organizations that want to create value quickly from an AI deployment. Let’s raise a glass and a plate of haggis to the Scot who is still helping us solve labor problems today. Connect with us: LinkedIn , Bluesky , X , Website To learn more about the services we offer, please visit our product page. Nick Reese is the cofounder and COO of Frontier Foundry and an adjunct professor of emerging technology at NYU. He is a veteran and a former US government policymaker on cyber and technology issues. Visit his LinkedIn here . This post was edited by Thomas Morin, Marketing Analyst at Frontier Foundry. View his Substack here and his LinkedIn here . Subscribe now Leave a comment