Agentic AI and the Rumors of SaaS’s Demise

Introduction: A Conversation About Big Shifts

When I joined Tobias Macey on the AI Engineering podcast recently, we covered a lot of ground — everything from whether SaaS is truly “dead,” to what is required to take  AI agents from prototypes to trusted business tools. In this post, I want to give you a quick preview of a few of the ideas we explored. (Spoiler: it may not be the version of the future you’re hearing everywhere else.) If these ideas catch your attention, I’d encourage you to listen to the full episode — we went a lot deeper into what agentic AI really means for the future of software, business operations, and SaaS itself.

🎧 Listen to the full conversation here → 

First, What is  Agentic AI?

We kicked off our conversation by talking about what it means for something to be ‘agentic’. For me the term “agentic” can be applied to  AI systems that exhibit three key characteristics: (1) an ability to interpret tasks, (2) an ability to reason through steps and (3)  an ability to act with some degree of autonomy. In my view, that autonomy of action that characterizes the third step sits on a continuum that ranges  from somewhat rigid with quite proscriptive workflows through to fully autonomous agents. In my opinion an agent doesn’t have to decide on all steps every time, in order to be agentic. In fact, I believe that many use cases will limit  levels of autonomy and require more human oversight or proscribed workflows. 

Is SaaS Dead? 

We discussed the recent headline-grabbing claim from Microsoft CEO Satya Nadella on the BG2 Squared Podcast that all business logic, currently found in SaaS applications, will migrate into AI agents. In the episode, I shared my belief that SaaS isn’t going anywhere—but it is evolving. We talked about how SaaS platforms are starting to embed intelligent agents, how user interfaces are shifting from clicks to conversations, and how business models may move from seat-based pricing toward usage-based structures.

In other words, I believe  that Agentic AI will change SaaS, but I don’t think it will erase the need for platforms that deliver secure, reliable services and data. I see SaaS morphing into something more dynamic and AI-enhanced, not disappearing.

Fully Autonomous Agents? Not So Fast

Another theme Tobias and I got into was the realism gap around AI autonomy. In my view, while today’s AI models are impressive, they still hallucinate—meaning they sometimes produce outputs that are inaccurate or unpredictable. And that hallucination may not be a bug, but rather an inherent characteristic of how these systems generate and reason.

As I shared during the podcast, this leads me to believe that human oversight will likely remain essential for most meaningful business processes—at least for the foreseeable future. Therefore, rather than replacing people, agentic AI is most likely going to continue to augment human abilities in many enterprise use cases. Further, we are likely to move toward a world where humans supervise, validate and collaborate with AI systems, not hand over control entirely.

In other words, if you’re picturing a future full of unsupervised AI employees running your business—it might be time to recalibrate those expectations.

Why Traditional Automation Still Matters

I especially enjoyed our discussion around  how agentic AI fits alongside previous waves of automation methods like Robotic Process Automation (RPA). There’s a tendency to assume that agentic AI will make technologies like RPA obsolete, but in my opinion, the reality is more nuanced.

During the episode, I described the idea of a “software lasagna,” a concept I borrowed from my colleague Nahim: in software we tend to add new layers, but old layers don’t necessarily disappear. RPA may still have a role where deterministic, structured workflows are needed. Agentic AI can layer on top, using RPA as a tool, and adding flexibility and reasoning when and where it’s valuable.

In my view, the future isn’t a clean break from the past—it’s an increasingly rich stack of technologies working together.

Wrapping Up: Join the Conversation

These are just a few of the themes we touched on—the full conversation goes much deeper into the practical challenges and exciting opportunities that agentic AI is unlocking across the software world.

If you want to hear my perspective on what’s happening (and what’s not) with agentic AI and the future of SaaS, I hope you’ll give the episode a listen.

🎧 Catch the full conversation here →

Would love to hear what you think.