Selling Software vs. Selling Work

One of the most interesting aspects of AI and how it'll change the workplace is how it might disrupt the large SaaS players and the SaaS model itself. The SaaS business model typically is set up around licensed "seats," where a SaaS company sells an individual seat to an individual employee to give them access to the software. The idea is that the software makes the employee more productive, and that gain in productivity is 2x to 5x the cost of the seat. This formula has led to the emergence of thousands of successful SaaS companies and hundreds of billions of dollars in market cap.

AI changes this quite a bit. Unlike SaaS, AI doesn't just make an employee more productive. It does the actual work of the employee.

As an example, if AI could triage patients coming into a hospital and connect them to the right next step (a specialty referral, an inpatient admission, a prescription, etc.), that's not making the healthcare provider a little more productive, that's opening up a nearly unlimited capacity to triage patients because the Ai can do it significantly faster 24 hours a day. 

This is very different than making an employee 20% more productive. And it's different from replacing an employee with a machine. It's taking ownership of a work product and removing caps on output. Try to imagine doing this across multiple areas of work inside of a company. All kinds of limits to growth and productivity will be removed, and companies could be infinitely more valuable. 

Coming back to the notion of SaaS and selling seats. The AI company would never sell this way. The ROI isn't centered around increasing employee productivity, it's centered around doing jobs with nearly uncapped output. It'll be interesting to see how this impacts the SaaS business model that has become so prevalent over the last several years.

What Artificial Intelligence Is Not

With all the hype around AI, I’ve seen the media and investors get a little bit sloppy with the definition of AI and have been calling things AI that aren’t AI. When there’s a big trend like this one, there’s a temptation to attach things to the trend that shouldn’t be attached to the trend. I’ve heard it’s hard to get meetings these days with top VCs if you’re not talking about AI in some form so founders are certainly incentivized to stretch. It’s worth defining what it actually means and what things are software versus AI.

Software that is not AI is running off of mid-20th-century technology. It effectively acts like a digital calculator. A programmer codes the software to do something, and the software behaves accordingly. The software is not thinking. It can’t do anything it wasn’t coded to do. If the software makes a mistake, that is the fault of the programmer who coded it.

Ai is much different. Ai mimics human intelligence. It can do more than what it was coded to do. ChatGPT describes itself this way:

The ability to learn, adapt, and generate contextually relevant responses based on patterns in data is what elevates it from traditional software to the realm of artificial intelligence. It's a software that, to some extent, can exhibit intelligent behavior and respond dynamically to various inputs.

One way to think about this is that software makes the human mind more efficient while AI mimics the human mind. From an investing perspective, this difference is really, really important. As this technology develops, given the massive productivity gains from a technology that can do the job of a human infinitely more quickly, 24 hours a day with no breaks could cause us to totally rethink how to value companies that create and/or leverage this technology.