Replacing SaaS with AI agents
2024-12-23 15:35:00 +07:00 by Mark Smith
There’s a ton of talk these days about how AI agents are going to affect coding. We’ve all seen how these LLMs can speed up development by producing code rapidly and developers have been successfully increasing their productivity using them as sort of copilots. I think that makes sense. You’re still essentially writting code, and code that actually runs on the server.
But I’ve noticed many on tech podcasts, including the All-In podcast, the This Week in Startups podcast, and Keen and Tear's weekly tech roundup podcast, all talking about how SaaS products will be completely replaced by AI agents. Assuming that it’s technically possible, and I’m not convinced it is, how can this possibly be a good idea? Replacing a deterministic, 100% repeatable, inspectable and debuggable process, with a stochastic probabilistic black box that might suddenly decide to hallucinate or that it just can’t and will instead lie to you. Have we learnt nothing from the thousands of supply chain attacks from the past few years?
Replacing SaaS products with AI agents surely has to be the most unsafe thing you could possibly do. Sure there might be some apps where it won’t matter all that much, but most serious apps require some amount of privacy to be useful, and many could be turned into dangerous products because these AI agents could essentially be increadibly good man in the middle machines.
If you can’t be sure what code is running, then not only can’t you really be sure it’s going to do what you want, but you won’t be able to see what it did when something unnexpected does happen.
Of course as a SaaS developer myself I’m somewhat conflicted on this. I like the idea of rapidly developing some personal tools myself, but I honestly don’t see how they could be useful outside of personal hobby projects. This notion that SaaS will and should be replaced by AI agents is like saying cars should be replaced by Jello.
I will concede that Jello AI is a pretty cool sounding name for a company though.
You think tech is bad now, run by megalomaniac billionaires? That will seem hilariously quaint in a world where AI agents are mediating everything. Not only do AI agents make possible glitches and bugs that we can’t even dream of, but it might make possible ones that we literally can’t even imagine. I’m guessing that likely will not be everyone’s cup of tea. #