Should We Still Build SaaS Apps in the Age of AI Clones?
I gave another talk last night at our meetup, and a question came up about Adobe and whether we should even continue building apps — especially software as a service (SaaS) style apps.
I thought this was an excellent question, and I also believe that even though there is a lot of hype around AI being able to build clones in Rust or whatever language you choose, I am not that concerned.
Who am I going to trust?
First of all, a lot of what I look for in SaaS products is a company I can trust. With everyone building clones, who am I going to trust? Sure, the price might be compelling, but what about all of the software governance and security best practices that have been thrown out the door? Even if the person directing has some software experience, how rich is it? Does it equal the amount of planning sessions and code reviews that Adobe has spent building their products?
AI can build a feature, not the reasons behind it
Don’t get me wrong, I am super pleased that AI can do a lot of what we are seeing. But if you truly read the fine print, it has not reached feature parity, and I don’t think it ever will unless you have developers who understand what those features and technical constraints require.
Just because the AI can build a feature, it doesn’t know the design decisions behind that feature or why one approach can cause havoc in the future. Neither will the person sitting there telling the AI to proceed. That comes from experience and knowledge that makes a person wise in the way of development. AI still cannot do that.
It has improved a lot, but it still doesn’t see the big picture, nor does it understand the subtle decisions from a design session that help improve the product. It needs to be coached and corrected along the way to get things “right”. This is not something I see right now.
Can a clone reach feature parity?
Fast-forward to building a clone of one of Adobe’s products: is it possible to do this and reach feature parity? In my honest opinion, no. It can get close, but there are always subtleties that are just not in the code or design documents that AI will ever see. I am actually comfortable making this statement for all mature software that goes to production.
You need expertise in the form of human beings to wrangle the Wild West that has become our current landscape. AI has only made it worse, and it will continue to be messy until we all agree that even excellent AIs need experienced developers to help them navigate the waters and ensure governance and best practices.
Don’t let the hype get to you
Most of what I see is branding hype to try to get you to click on the latest scare tactic. What I have inside my brain, and what I have been building for years for clients, has become an art in some way. It isn’t just requirements. It is the art of translating what a customer truly needs, and then correcting them when they don’t actually know what they truly need.
An AI, for the most part, is a “yes” man. It will do what you ask, and mostly only what you ask. Sure, you can get creative with your prompts, but that requires skill and experience.
Keep building what you are building. Keep debugging. Keep honing your skills. There will always be a need for our expertise and training. Don’t let the “hype” get to you.
Just my two cents…
More on working with AI without letting it drive: AI Is an Eager Puppy. Keep It on a Leash.