One of the recurring arguments in discussions about AI is that models are becoming commodities.
A few years ago, access to advanced AI models was limited to a small number of companies with enormous research budgets. Today, developers can choose from a growing list of capable models from providers such as OpenAI, Anthropic, Google, Meta, and others. New models appear regularly, costs continue to fall, and switching between providers is becoming easier than many people expected.
As a result, simply having access to a powerful model is becoming less of a competitive advantage.
That does not mean data has become irrelevant. If anything, it may have become more important. The difference is that the role of data is changing.
For years, the technology industry repeated the phrase that “data is the new oil.” The implication was straightforward: collect as much data as possible and value would eventually emerge from it.
In practice, most organizations discovered that vast amounts of generic data are not particularly useful. Storing information is easy. Extracting insight from it is much harder.
What seems to matter more is having access to data that is closely connected to a specific workflow, decision process, or business problem.
When I think about products I am building today, such as Founder or Blest, I do not spend much time wondering how to collect massive datasets. I am more interested in understanding what users actually do.
Which recommendations do they follow?
Which screens do they ignore?
Where do they abandon a process?
What information do they repeatedly search for?
What do they change after receiving a suggestion from the system?
Those interactions generate a different kind of data. They are not merely records stored in a database. They are signals about how people work.
That distinction feels increasingly important as AI becomes embedded inside software products.
Many AI applications look remarkably similar from the outside. They all connect to comparable foundation models. They all provide chat interfaces. They all generate text, summarize information, or automate tasks.
The model is often the least interesting part of the product.
The more interesting questions are what context the application understands, what workflows it supports, and what it learns from repeated usage.
A scheduling application can learn how a particular team plans projects.
A financial platform can learn how an organization reviews spending.
An educational system can learn how students progress through coursework.
A product like Blest could eventually understand how institutes manage classes, teachers, attendance, and communication with students. That accumulated operational knowledge may become more valuable than any individual AI model integrated into the platform.
The same pattern appears across industries.
As foundation models become widely available, the differentiator increasingly shifts toward proprietary context and workflow knowledge. The organizations that understand their users best gain an advantage that is difficult to replicate.
This is not a new idea. Software has always rewarded deep understanding of a problem domain.
What may be changing is the speed at which that understanding compounds.
Every interaction generates data.
Every piece of data improves context.
Better context enables better recommendations and automation.
Better experiences attract more users.
The feedback loop is familiar, but AI appears to amplify it.
That is why I suspect the next generation of successful software products will not necessarily belong to the companies with the largest models.
They may belong to the companies that learn the most about the work their users are actually trying to accomplish.
Related Sparkio Products
AppGrid Platform designed to help independent developers connect their products with potential customers.
Founder Interactive business history simulator inspired by the stories and decisions behind the world's most influential companies.
Blest Platform for managing language institutes, students, courses and learning operations.