"A SaaS is just a database wrapper"

"A SaaS is just a database wrapper" sounds funny but there’s something to it.

Obvious exaggeration, but it reflects something that is resurfacing as LLM based applications in general (and AI agents in particular) become ubiquitous: the need for new interfaces to interact with data is declining.

In simple terms, a SaaS is a cloud app with an interface to access a database in a specific vertical (say, dry cleaning orders or tasks in a project).

Instead, most new AI apps use natural language to get to the same result.

When you learn to use a new piece of software, what you are doing is translating a set of rational instructions (mostly based on language abstractions) in your head into clicks and actions on a dashboard designed by someone else, resulting in some data saved somewhere.

It feels like an unnecessary step we could skip and finally move past most interfaces.

I'm seeing this myself when developing AI agents to solve business problems.

In most cases, the best agent is simply the one you can talk to like you would to someone on your team.