Is AI Going to aSaaSinate the Cash Cow?

Or, how AI is poised to disrupt the SaaS market

In recent months there has been a lot of discussion about the impact of AI on the Software as a Service (SaaS) market, with many experts predicting with increasing certainly that AI will kill SaaS. I’m not sure that I can see this as a realistic outcome, but I do think that AI is going to significantly disrupt the SaaS market.

The current SaaS model relies on finding a problem and figuring out how much users would be willing to pay to solve it. This model has worked well for many years, but it’s based on the assumption that customers are not in a position to easily solve this problem for themselves. With the advent of AI this assumption is being challenged by the ease of generating working code with little more than a simple description of the problem. This combined with the increasing availability of low-cost cloud infrastructure means that users can now build and maintain their own DIY services at a fraction of the cost of a SaaS subscription.

On the face of it this seems like the start of the end for Saas. Who in their right mind is going to pay $20 a month for a SaaS service when they can generate their own DIY code for (almost) free and host it for a few dollars a month? Personally, I think that this is a misunderstanding of both the value that SaaS provides and the limitations of AI generated code.

While I concur that AI can generate code to solve a given problem, this code is unlikely to be complete or “production ready” (especially if you’re not already an expert). And as easy as it is to deploy AI generated code to the cloud, operating a service requires both expertise and ongoing effort. This means that once you’ve generated your own DIY service you’ll need to make further investment on two fronts; firstly to prompt the AI to generate more complete and robust code, and secondly to operate and maintain the service over time. What was initially a low-cost solution quickly becomes expensive in the time required to maintain it (who’d have guessed).

Now I know that some of you will be yelling at the screen telling me that AI is only going to get better. While I agree with this sentiment; I think we’ve already seen the rate of this improvement slowing, and better isn’t necessarily smarter. AI is very good at generating code to solve a given problem, but it is much less able to understand the broader context or design the solution for future changes. This means that while AI can generate working code (sometimes you may need to squint) it’s unlikely to generate code that is maintainable, scalable, or secure. This problem only gets worse as AI changes the codebase without an overarching architecture or a foundation to support the new code; leading to a fragmented and inconsistent codebase that is difficult to maintain (again mitigated if you’re an expert).

And this is where SaaS steps back in with the same argument that it has always had; wouldn’t you rather pay someone else to do the boring work so you can just solve your problem? Now I’m guessing this is also where you’re back yelling at the screen that I’ve not actually made a point yet? Well I’m getting to it, I just needed to set the scene first.

Drum roll please…

My conclusion is very simple; the calculation that SaaS providers have been using to figure out what users will be willing to pay for their services is going to have to significantly change. Where there was once a reasonably high activation energy requirement to go DIY, there is now a much lower barrier to entry due to the power of AI. We can think of AI as the DIY service catalyst, it wont kill SaaS but it’s going to make SaaS customers much more price sensitive. Is this a bad thing for SaaS providers? I don’t necessarily think so, but it looks like a significant market disruption that will require SaaS providers to carefully rethink their value proposition.