AI Redefines the Software Landscape: Towards an Era of Transformation

AI and the Threat to Software Publishers

Artificial intelligence will not annihilate all software publishers, but it challenges solutions that lack business depth. Value is shifting from mere lines of code to mastery of complex business rules. Historically, publishers have been protected by the complexity of their code, but AI now offers alternatives that replace once-costly functionalities. However, not all solutions are exposed in the same way. According to Benoît Wambergue, Vice President of Cloud at Fortero, the risk of substitution heavily depends on business depth and the structure of the data being processed.

A Threat to Standard Tools

The concept of "SaaSpocalypse" illustrates the planned obsolescence of software by AI, primarily targeting low-value systems. Financial software and basic CRMs, which often rely on standardized processes or simple data flows, are on the front lines. Today, a powerful language model can easily infer these rules, rendering these tools obsolete.

Exposure to this threat varies depending on the nature of the system. "Engagement systems," which are highly dynamic, differ from "record systems." The latter collect key business data, such as industrial nomenclatures or cost structures. The more a software processes raw and unstructured data, the more likely AI can replace it. Conversely, a complex functional business depth acts as a shield against this threat.

"AI-Coding": A Deceptive Accessibility

Generative AI democratizes development through "Vibe Coding," allowing functional software to be produced without mastering syntax. However, this apparent accessibility hides a structural trap. Language models often produce verbose, clunky, and heavy code. Without expert human supervision, these programs can become technical dead ends at the first bug.

The risk lies in the loss of intellectual control over the tool. LLMs have a limited context window when faced with applications comprising millions of lines. They struggle to grasp overall interactions and the risks associated with deep modifications. A team that no longer understands the internal workings of its code loses its maintenance capability. This reactive drift leads back to the era of makeshift "homegrown solutions," once managed by isolated system administrators. These hybrid systems, blending AI and trial-and-error, create colossi with feet of clay. Ultimately, a clear divide will emerge between seasoned developers, augmented by AI, and mere executors producing fragile software.

Human Expertise in the Face of AI

In the face of computational power, humans retain a unique value: discernment. AI responds to requests without questioning their validity. This lack of filtering inevitably leads to "functional inflation," which burdens systems without creating value. In the industrial sector, this nuance is crucial. A client often requests a feature out of habit rather than strategic necessity. The expert must filter these needs to avoid unnecessary functional inflation. Humans know when to steer clients toward a more efficient industrial standard. Their ability to say "no" protects the long-term stability and performance of the solution.

The seasoned developer cannot be replaced. They use AI as a productivity lever while remaining the guarantor of the overall structure and business sense. "The added value of a human is to challenge the client's request. An LLM will never say, 'no, I won't do that.'"

Towards a New Software Economy

AI is challenging the economic foundations of the software publishing sector. User-based payment is no longer an appropriate metric. "Payment per user is probably no longer an appropriate metric. We will have much more autonomous software." Publishers must then shift towards pricing based on value or actual consumption. The operating cost no longer depends on the number of connected humans but on the computing power required. This transition is crucial to reflect the actual performance delivered to the client company. The market is still seeking its new balance between token costs and business benefits.

To avoid disappearing, publishers must transform their approach. The exercise involves attempting to solve their own business problem solely with AI. If an orchestration of agents replaces their value proposition in just a few clicks, the risk of substitution is total. The dependence on the survival of business intelligence is already crystallized in the software. Processing large volumes of structured data remains an asset for traditional software. A complex industrial calculation, like MRP, requires a comprehensive view of the production chain. AI still struggles to manage these levels of nomenclature without generating prohibitive computing costs. Therefore, software must focus on these areas of expertise where generic tools fail.

"We are trying to reframe the problem and solve exactly the same one using only AI. If upon your arrival, the product will be replaced within 5 years."