Software is Eating Labor

If there's one key takeaway I can offer today, it's that the labor market, while almost obvious, is significantly larger than the software market. The worldwide SaaS market is approximately $300 billion annually, with a global market cap of about $2.2 trillion. In contrast, the U.S. labor market alone is valued at $13 trillion. This doesn't imply the software market will suddenly reach $13 trillion a year. However, the labor market is now the significant prize that software is targeting.

My thesis is that almost every software company has essentially taken a filing cabinet and transformed it into a database. This innovation accounts for the $2.2 trillion in market capitalization and $300 billion in annual software revenue. A prime example is Saber Systems, a joint project between American Airlines and IBM, back when IBM was the world's leading computer company.

I mention all of this because efficiency hasn't significantly improved. Filing cabinets were read by humans, and digital records are also read by humans. For example, this woman, assisting a customer with support needs, would have once looked at a piece of paper. Now, she looks at a computer. This is crucial to understand because the entire business model of software must change. It absolutely has to change.

Zendesk is at a critical juncture. Their revenue could either drop to zero or triple. The outcome is uncertain, even for them. I've spoken with their CEO, and they are piloting outcome-based pricing in New Zealand. So, New Zealand might hold the answer to our questions; stay tuned. This illustrates how much larger these labor or quasi-labor markets are. The $13 trillion in wages we discussed dwarfs software revenue.

AI handles intermittent demand, which is a significant advantage. Consider a Black Friday retailer: sales spike dramatically around that time. This would typically require hiring numerous cashiers or, for online retailers, many customer support staff to handle queries. What happens in January? Do I fire everyone, only to rehire them in November? In reality, I'd need to hire them even earlier, perhaps September, for training. This presents a tricky situation.

AI isn't bothered by demoralizing jobs. It offers excellent regulatory certainty. For example, I co-founded a company called The Firm. We received quarterly training on UDAP: Unfair, Deceptive, and Abusive Practices. Many laws govern what you can and cannot say to a customer. Imagine calling a customer to say, "Hey, you owe me money." The customer responds, "F you." If the person on the other end is having a bad day, they might retort, "F you back, customer." They would then face consequences for that.

My favorite example involves someone who studied languages. I speak Russian, Japanese, and a smattering of Spanish, having spent a significant amount of time learning them. It's a fact that an AI nurse could be incredibly beneficial. For instance, what if I only spoke Farsi? Does Stanford have Farsi-speaking nurses who can discuss my pain level? What if I only spoke Mongolian? Now, an AI nurse, an AI collections agent, or an AI negotiator could handle these situations. All these functions can be performed instantly in dozens of different languages.