Editing The AI agents are coming for SaaS

[[file:The_AI_agents_are_coming_for_SaaS.jpg|500px]]

There are times when a hypothesis is more than just another statement, and becomes a mirror reflecting the age we live in. Jeremy Blaze, after years designing SaaS applications and establishing himself as an authority in the field, now argues that these systems will disappear with the rise of AI agents.

It might sound dramatic, but his assessment resonates with recent Microsoft statements, which likewise predict the decline of SaaS due to the emergence of agentic systems. In this article, I aim to explore that premise through a reasonably rigorous lens, while also offering a hands-on perspective rather than a purely theoretical one: how feasible is it that agentic systems will replace SaaS as we know it?

After more than a year working with agent-based systems, I would say the answer is yes. These agents move us closer to the theoretical ideal of a Gödel machine, a system capable of self-reprogramming efficiently after proving its next version is superior. This is no longer science fiction, but rather a natural outcome of building agents that can adapt, learn, and improve autonomously toward a defined goal.

Blaze’s argument isn’t conjured from thin air; it stems from deliberate and challenging practice. He’s evolved traditional SaaS interfaces — from form-based and dashboards to CRUD flows (Create, Read, Update, Delete) — into increasingly collaborative environments where agents act as co-creation partners with humans. He has built internal tools that interface with years of accumulated knowledge via natural language, accelerating the resolution of complex challenges with precision and speed. His approach is not theoretical — it’s fully operational.

Meanwhile, from its throne atop the software world, Microsoft, under the strategic leadership of Satya Nadella, insists that SaaS platforms are essentially CRUD databases with business logic — logic that will soon live in an AI layer where agents manage multiple repositories without rigid interfaces. These are not just opinions: a study by AlixPartners shows over 100 software companies are being displaced by generative agents that operate without dashboards or predefined data hierarchies. In the market, evidence abounds: companies like Klarna are dropping SaaS solutions like Salesforce and Workday in favor of internal agentic systems or solutions from emerging vendors. Some CIOs suggest this shift will be substantial, although many IT teams remain skeptical and frame it more as an evolution than a full replacement.

This transformation, however, won’t be painless. Despite the revenue potential — forecasts estimate $52 billion by 2030 in the agentic systems space — the business model remains vague and tough to implement, especially in regulated or financial environments. Microsoft’s own report advises a phased transition: agents first as “assistants,” then as “digital colleagues,” eventually managing entire workflows with human oversight under a new role: the “agent manager,” or even reaching the point where agents become the managers themselves and supervise humans.

And if we’re talking paradoxes, there’s a poetic irony in the fact that Blaze, a veteran who helped solidify SaaS interfaces, has now arrived at this conclusion — and he does so from within. With some practical experience of my own in developing and deploying agentic systems, I can modestly confirm: agents are not just powerful — they are edging closer to a form of self-reference, a nearly Darwinian loop of continuous improvement. They can rewrite their own logic, albeit imperfectly, through iterative learning. The ramifications are far-reaching: who owns the behavior of such systems? Can we continue training agents without losing control? How do we relate to entities that keep refining their own rules?

For example: an agent managing a fleet of drones, optimizing routes and learning from weather patterns — without a graphical interface. Or a financial agent automating credit decisions and adjusting its risk model on the fly as it identifies new patterns. Or a design tool, like the one Blaze built, extracting references from natural language and evolving its internal corpus based on each project. In all these cases, the traditional SaaS layer — with its screens, dashboards, and rigid flows — is irrelevant, or at least secondary.

Blaze’s assessment isn’t a doomsday prophecy but a grounded proposal: agentic systems don’t just shift the landscape — they redefine it. Will SaaS be phased out before 2030, surviving only as legacy infrastructure in companies increasingly burdened by technical debt and strategic disadvantage? Data, experience and expert analysis seem to align toward a point of no return. Yet this very shift — paradoxical and fascinating — opens up a new set of ethical, governance and design questions that will require serious attention and oversight.

Read the full article here: https://medium.com/enrique-dans/the-ai-agents-are-coming-for-saas-2706a5fde48d