Everyone could build a SaaS before, it just is quicker and cheaper now because you don't have to hire a coder. The difference between success and failure is still building the "correct" product, for both meanings of "correct". For those who are unaware, meaning #1 is to build the product that the market will adopt, meaning #2 is to build the product correctly so it is sustainable and maintainable.

LLMs do not guarantee either of those, so unless they improve to the point where they do, this really doesn't change the overall market, it just changes the level of resources required to play.

What LLMs are going to replace is more likely to be the spreadsheets that departments run on, sharepoint and salesforce lists, and other such low-cost, low-effort, and legacy tools that departments tend to use for their own local processes.

You can build the correct product. The bigger issue is convincing people to take a chance on some unknown vendor. Besides no company or any size wants to or should depend on a lot of unsupported vibe coded internal slop. It’s no different than the old VB apps or Access apps

Anyone who thinks Salesforce is easy to replace hasn’t been part of large sales and consulting organizations - or hasn’t been close to sales side. Heck, even AWS’s internal consulting department depends on Salesforce.

> Will value shift entirely to AI platforms and infrastructure, with most SaaS becoming commodities?

At some point AI platforms will struggle to monetize at the model API layer due to local models catching up. This is why the model providers are focused on AI model integrations at the application layer which invades traditional SaaS companies.

You now need to think very carefully about 'what' to build and assume that someone with a AI agent can clone your SaaS and race you to $0 as a risk.

When building becomes easy, curation becomes the moat.

Anyone can spin up a SaaS now, but knowing what information actually matters to a specific audience is still hard to replicate. That's domain expertise, not technical skill.

I think the value is shifting from "can you build it" to "can you filter the noise." The bottleneck isn't access to information anymore - it's attention.

Attention is all you need?

It is a bit weird watching the industry pivot from ‘no-code’ to AI agents. But to be honest, i think if a prompt can knock out a bespoke app in five minutes, the traditional SaaS model is going to be as relevant as a fax machine, leaving the real value in whomever owns the pipes and the trust to actually run the stuff, and maintain it, once we've pivoted to the next shiny paradigm shift.

Ok, anyone can build it - let’s say I believed that.

Now try to displace Salesforce, Slack, ServiceNow, Workday, etc.

No corporate buyer is going to take a chance on your homegrown vibe coded replacement that they saw on HN where the author didn’t even seem to know the importance of having SSO support.

Besides “no one ever got fired for buying IBM”. Why would I take the reputational risk of not just buying from the well known company who is already at the right square of “Gartner’s magic quadrant”?

Writing software has never been the challenge. If anyone can write software, Microsoft can just throw a few engineers at it and bundle it in thier corporate seat licensing

Large software platforms with their own ecosystems will be spared this fate. Mostly because AI can only create small and trivial apps and further model improvement in context size or otherwise may not be forthcoming so easily.

So large software platforms, think Jira/Confluence, MS Teams, SAP etc are not affected. But AI will definitely eat the solo dev SaaS, especially those handling trivial use cases.

Support. The boring part. Having a person that can make reasonable decisions for the customer.

Some of this can be automated. We’ve seen the result with the big companies though.

Everyone can’t build SaaS and sell SaaS! It’s hard to sell to get distribution

Neural networks are not the magic wand that will create a high-quality app for you.