(This was my thesis when I first used ChatGPT 3 years ago)

In the last three decades we have seen SaaS dominate. Software has this magic property where it takes something physical and makes it infinitely scalable and reproducible. SaaS companies took a huge profit/premium as they de-physicalized so many things. While this trend will continue, a lot of SaaS companies will suffer as AI will make those products obsolete. Why buy a software package when you can have an AI create the software for you instantaneously, and it will be more customizable for you. 

Going forward, you have to offer something physical, a commodity, something licensed, or a network to be valuable. Oh, and you still need to be a master of software to digitalize as much as you possibly can, all just to stay relevant.

I think this is really big. Just as software de-materialized the physical world, and created huge advantages to software, I think AI will de-materialize (in a sense) the creation of code. Code will become dirt cheap, just as most digital goods are dirt cheap. This will be very good for most going forward, but it will be a big shift. Physical and privileged goods (things that require a license to sell, like banking) will be the things that have value. Pure SaaS is gone. Get out.

More people see this trend now than ever have. I have told this thesis to others who didn't believe, but now they do. There are companies/products that sell this already. I am told about them all the time, and Ill probably start using them when I start my startup software company next week. The rate and pace of change is pretty startling. I have even heard estimates that as much as 80% of coders are inferior to AI coding agents.

Now, what is my more specific outlook on software.

I think we are a long way off from anyone just speaking software into existence. I think it will come, but it is unproven that the rate of change large language models are making will be able to keep that pace. Also, and this might be bigger, people and companies take a lot more time to adopt software and technology than we think. I think we should prepare for it, however, it may take a while to play out. Maybe I am wrong, and it will not. The timing is uncertain.

Next, I think we will see, right now, a wave of one person or small team start-up companies, like what I am doing, that are going to build software and sell it. We are going to see a flood of options. The trick is can they market it. Anyone with a half decent idea can build a product now. This, more than anything, will make us see a lot of competition in SaaS and drive prices down to small and minimal costs to use SaaS.

Now, I think there are two paths. It is not clear. One is that software is so cheap so fast, that for the most part, people dont want to build their own software. If there is a free open source version, or there is a super cheap private version. You would rather use this than trust your employees to build their own version. The second path is what I laid out above.

Something that is a really big reason that companies will not build their own software is that they need a third party, SOC audited, software. It is too much of a security risk to build it themselves, or it is still too much work internally to validate as a secure product. This is going to be a bigger deal in any industry that is more security minded.

To expound on an earlier point, companies talked a long time to adopt new technologies. This is really true for companies that have an existing solution. Banks and governmental agencies run on super old code. At one point they were early adopters of software to make their processes much more efficient. Now they have very old  and clunky code and have not built new more efficient software. The decision to not build new codes is more a factor of cost and risk vs the benefit. It would be very costly to replace code that works, even if the new code could be a little better. Additionally, it is not worth the risk that replacing the code could potentially break the software and distrupt existing processes and client experiences. That risk is not worth the switch.

So, to clarify, large companies and governmental agencies will probably stay on whatever code base they are using. If their audits are dependent on a certain code base, or them using some third party softwares, they are golden. Smaller companies, or software that is not as mission critical, those might be swapped out for something new and better. Maybe the company will build their own. Maybe they will buy something super cheap from a solo entreprenuer. Time will tell, but I don't think it will be as black and white as I saw before. 

Regardless, I don't think SAAS will see the killer valuations we once saw. I think we are going to see software start to be commoditized, but not killed. There will be more options and more competition than ever, and prices will fall down to cover costs. Eventually, maybe AI will kill SaaS, but that is far out and unproven. If AI is good enough to kill SaaS, it is taking a lot of other things down with it.