I’ve been building software products for 30+ years. It’s what I love to do.

Life got a lot better for software developers when the web first came along and we could use it for distributing software over a browser. This eventually led to the rise of what we call “SaaS” now — Software as a Service. The idea is simple, but powerful. Instead of paying for traditional software that you buy and install on your own hardware, you gain access to software “as a service” (accessed over the web).

The popularity of SaaS was somewhat predictable. Customers want to accomplish goals with software, and the lower-the-friction to get that benefit, the better.

As the old adage goes, “People want holes not drills.”

In the same way, all things being equal…

And so, the natural evolution, for many use cases is to go from Software as a Service to Results as a Service.

I’m not saying SaaS is going to die, what I am saying is it’s going to evolve.

And that’s where agents come in. (You didn’t think I was going to get through an entire post without talking about AI agents, did you?)

At their simplest level, agents are A.I. powered software that can complete goals that require multiple steps. They basically deliver results.

So, where software is headed is that more and more capability will be delivered via agents that produce results. Some agents will be simple — almost trivial (I’m a believer in the Minimum Viable Agent), and others will be super-sophisticated involving a collection of agents and a way to orchestrate their work to accomplish a goal, often autonomously. This is popularized by agent frameworks like CrewAI.

The example I like to use (because I wrote it) is the Company Research Agent that’s part of agent.ai. It does exactly one thing: given a company name, it will produce a pretty detailed report (the “result”). It goes through a couple of dozen steps using a variety of LLMs (GPT-4o, Claude, Google Gemini) proprietary data sources to produce its output. I built it for myself, and use it every day.

Here’s a sample report on OpenAI.

I think you’ll be amazed at the kind of data and insights it can produce. Things that were simply not possible before the age of generative A.I.

Note: Agent AI lets people earn free credits on the system when users sign-up with a shared report link like the one above. But, I’m not doing it for the free credits, I promise). 🙂 

One could argue that AI agents and RaaS generally are really just SaaS. And one would be right. Just like Software as a Service is really still software, Results as a Service is really still SaaS. What changes is what’s being purchased. With SaaS, you’re purchasing access to software. With RaaS, you’re paying for an actual result. Aaron Levie from Box calls this “Outcomes as a Service”, but I prefer RaaS, because it’s easier to say out loud than OaaS. The other clever take I’ve seen is Service as a Software (SaaS). Which is clever, but confusing, since the acronym doesn’t change. Similar idea though.

I’m looking forward to this next stage of evolution in software. It’s better for customers, and ultimately, it’s better for software companies because they get even closer to creating the actual value customers want.