There’s a phrase that in recent months has become almost a strategic ultimatum: either you’re AI-native, or you’re dead.
I understand why it works. It’s fast, brutal, easy to remember. But for many mature companies, it’s also the wrong phrase.
The problem is that it describes artificial intelligence as if it were just another new technology to be incorporated into the product as quickly as possible. I see it differently: AI is a transformative discipline. Like electricity, like the railways, like the internal combustion engine. It doesn’t just improve what exists. It redesigns the system.
If that’s the case, then for SaaS producers the question isn’t how much efficiency can I extract, but how much more value can I generate for the customer. The point isn’t just automation. It’s augmentation. Doing more, at greater scale, with greater scope and ideally better.
This is where many companies risk misreading the situation. Because if they look at AI only as a lever to reduce costs, they will end up using a historic transformation as if it were a small optimization project. And in the meantime, someone else will use that same transformation to change the nature of their offering.
For SaaS vendors, this is the heart of the game. As models become more powerful, some software components will tend to commoditize. Writing text, generating code, synthesizing data, acting as a conversational interface: all important capabilities, but increasingly indefensible on their own. The risk isn’t that SaaS disappears. The risk is that it remains too small relative to what is happening.
The software that will be compressed is the one that keeps selling mainly screens, standard workflows and isolated features. The software that will grow is the one that controls what a generalist model doesn’t have and won’t easily have.
What? First and foremost, culture. That is, the way an organization truly decides, what it trusts, where it accelerates and where it slows down. Then process: not the theoretical process from the manual, but the real one, full of exceptions, escalations, waivers, constraints, responsibilities. Then trust, which in enterprise is always a decisive currency. Finally data: not abstract data, but data that is contextualized, governed, historicized, connected to workflows and therefore useful at the moment when a decision needs to be made.
This, for me, is the true growth agenda for SaaS in the cycle that is opening: enriching the value of the offering, not just adding AI to the offering.
It means moving from “I help you perform operations” to “I help you govern a system.” It means incorporating into the product domain ontologies, risk thresholds, approval logics, auditability, integration with existing systems, observability, supervision and continuity of service. The more intelligence becomes abundant, the more value rises toward what makes that intelligence reliable in the real world.
This is why I’m skeptical of the idea that, for a serious incumbent, the goal should be to become AI-native in the literal sense. That is a useful category for many startups. For those who have legacy, history, clients, reputation, tacit knowledge and trust accumulated over time, the right word isn’t to be reborn: it’s to transform.
Legacy is not just a burden. If it is healthy, it is an asset. The challenge is to prevent it from becoming an alibi.
This is why I continue to prefer a different formula: AI First, Human Led.
AI First means that artificial intelligence should not be treated as an accessory feature, but as the center of operational, commercial and product redesign. Human Led means that direction, judgment, responsibility and the relationship of trust remain human. Not out of nostalgia, but because that is where an increasing share of value will concentrate.
Moreover, the more capable AI becomes, the less sense it makes to use error frequency as the sole metric. Today we ask: how often does it make mistakes? Tomorrow we will need to ask above all: how much does it matter when it makes mistakes? Models will be increasingly less fallible in ordinary activities, but they will have more agency, more autonomy, more capacity to act on workflows, code, documents, customers and decisions. We will move from more frequent but small errors to rarer and more impactful ones.
This is why the difference, in the coming years, will not just be technological. It will above all be managerial.
Technology will be the necessary condition to stay in the game. But winning will be the companies that can better define work, delegate better, verify better, redesign better. Ultimately, this is the most interesting paradox of AI: the more intelligence becomes economical, the more valuable become the human abilities to direct it.
I see in this a very concrete possibility of a new humanism. Not a romantic return to man versus machine, but the opposite: people equipped with very powerful tools, yet called upon to exercise even more judgment, more vision, more responsibility.
The future of SaaS, then, is not a frantic race to the AI-native label. It is a profound transformation of the offering. Less software as a catalog of functions. More software as augmented capability. Less AI as a slogan. More AI as an industrial discipline.
And above all: less fascination with what the model can do on its own. More attention to what the customer cannot afford to leave without culture, process, trust and data.