Business is complex, and consequently, SaaS is complex. AI can help manage some of this complexity. As a result, people are using AI to build vertical SaaS solutions.
For example, consider a human who translates between two systems of record like Zendesk and HubSpot, or between two complex features within one system like Zendesk. A subset of that translation layer could be automated with AI. This represents the current trend in vertical AI SaaS: building another SaaS platform that serves as a translation layer for a specific vertical.
There's nothing wrong with this approach, except that it introduces the very complexity it tries to solve, over time.
SaaS becomes complex because as it grows, SaaS platforms are incentivized to solve all the problems of their entire customer base. Progressive disclosure can only do so much before a subset of customers concludes, "this is too complex and expensive."
The complexity stems from trying to serve everyone, and the translation layer is just another layer of complexity that's going to serve all users. While everyone wants a tailor-made SaaS for their business with some unique features, it's extremely expensive to build that. Therefore, the huge price tag of SaaS solutions is perfectly rational.
But what if we could make it cheaper? What if the very technology that builds the translation layer could build and maintain the SaaS itself?
If Klarna can build their own SaaS to replace Salesforce with this iteration of "AI," then the next iteration should hopefully allow smaller businesses to do the same.