Is artificial intelligence about to destroy the licensing model that ruled for decades?
For decades, the software industry built one of the most profitable business models in history: build a product once and sell the right to use it millions of times.
First came perpetual licenses. Then subscriptions. The way of charging changed, but the underlying economic principle stayed practically intact: someone builds an application, controls its distribution, and the rest of us pay to use it.
That model turned companies like Microsoft, Adobe, Autodesk, and many others into tech giants.
But something is changing.
And we may be witnessing something comparable to what the music industry went through with Napster in the late nineties.
The difference is that this time, nobody needs to pirate the product.
We can simply build another one.
When Napster changed the rules
Napster appeared in 1999 and put an entire industry in check.
Until then, record labels controlled much of the production, distribution, and sale of music. If you wanted to listen to an album, you usually had to buy the whole thing, even if you only cared about two songs.
Napster didn't invent music, and it didn't make it cheaper to produce. What it did was reduce the cost of copying and distributing it to practically zero.
All of a sudden, consumers could access millions of songs without going through the traditional commercial channels.
The industry reacted with lawsuits, restrictions, and fierce resistance to change.
But the real problem wasn't Napster. It was that technology had altered the economic conditions the business rested on.
In time, iTunes, Spotify, and new forms of monetization appeared. Music didn't disappear. Neither did the musicians.
What changed was how the economic value of their creations was captured.
The film industry went through a related transformation — first with digital distribution, later with streaming.
And now, more than two decades later, we could be facing an even deeper transformation.
This time, in the software industry.
Crafting Apps: a sign of what's coming
Recently a project called Crafting Apps appeared — a collection of free, open-source applications that aims to offer alternatives to commercial tools used in areas like image editing, video, audio, documents, spreadsheets, and design.
The project uses Rust and proposes an architecture that lets you interact with the applications through graphical interfaces as well as automation mechanisms, including interfaces aimed at AI agents.
Some of these applications are still in early stages of development. It would be wrong to claim they can already fully replace mature products like Photoshop, Premiere, AutoCAD, or Microsoft Office.
There's also not enough evidence to attribute any significant economic impact on those vendors' license sales to them.
But focusing only on comparing features would miss what's actually interesting.
The news isn't that another free alternative to a commercial application has appeared. The news is that the barriers to building those alternatives are changing.
Free software has been offering competitors to commercial products for decades. Linux, LibreOffice, GIMP, and Blender are examples of projects that have shown it's possible to build sophisticated software outside traditional licensing models.
Still, developing and maintaining those products has required enormous amounts of human work, communities of contributors, and many years of evolution.
Generative AI introduces a new variable into that equation: the possibility of considerably accelerating software development, testing, documentation, and maintenance.
And that changes the rules of the game.
The cost of building software is falling
Historically, building a complex application required a considerable investment.
You needed software architects, developers, designers, database specialists, quality engineers, security experts, and infrastructure teams.
It wasn't just writing code. You had to design the architecture, implement features, fix bugs, build interfaces, document, test, and maintain the product.
All of that is still necessary.
What's changing is how much human effort is required to do certain parts of that work.
Today, an experienced developer can use AI agents to generate components, implement features, build automated tests, analyze entire repositories, and even coordinate development tasks in parallel.
That doesn't mean a single person can build a perfect replacement for any enterprise product tomorrow.
It means part of the work that once required large teams can start being done by much smaller ones.
And when the cost of production falls, so do some of the barriers that protected established vendors.
Think of a commercial application that costs hundreds of dollars a year.
Its maker isn't just charging for the code. They're also recovering historical investments in research, development, maintenance, support, infrastructure, and marketing.
But what happens when a competitor can build a reasonably functional alternative for a fraction of that investment?
And what happens when it's not one competitor, but hundreds?
Competition stops being limited to companies that can fund large engineering teams.
Code starts to lose some of its economic scarcity.
And when something stops being scarce, it becomes much harder to charge for it the same way.
The real problem for licensing
For years, the software industry has perfected its ability to turn features into recurring revenue.
An application you once bought once can now require a monthly or yearly subscription.
The SaaS model — Software as a Service — took that logic even further, combining access to the software with cloud-hosted services.
From a business standpoint, the model has clear advantages: predictable revenue, continuous updates, and long-term commercial relationships.
But from the consumer's standpoint it has also created growing dependence.
We pay subscriptions to edit documents, store files, design images, manage projects, automate processes, and use tools that, in some cases, we only need occasionally.
The problem isn't paying for software.
The problem appears when the price of accessing a feature stops being related to the cost of getting an equivalent alternative.
Imagine a small company that needs a tool to generate reports, process documents, and manage an approval workflow.
Traditionally, its options were to buy a license, hire a SaaS service, or fund custom development.
The third option used to be the most expensive and risky.
But with AI-assisted development tools, that relationship could start to flip for certain use cases.
A specific application, designed exactly for the company's needs, could end up cheaper than years of subscriptions to a general-purpose product.
And then an uncomfortable question appears for the industry:
Why keep paying indefinitely for features we can now build and maintain at a reasonable cost?
The answer will still depend on complexity, security, maintenance, and the guarantees required. But the question no longer seems as absurd as it did five years ago.
From buying software to building solutions
Maybe the most important change isn't the appearance of free alternatives to commercial products.
Maybe it's the transformation of our relationship with software.
For decades we got used to adapting our processes to the applications available.
We bought an ERP and modified our internal procedures to fit how it worked.
We bought a CRM and reorganized our sales flows around its capabilities.
We hired a project-management platform and ended up working the way its developers had imagined.
The logic was understandable: custom software was expensive.
But what happens when that economic constraint starts to disappear?
We could be entering a stage where companies no longer have to constantly choose between adapting their processes to software or funding expensive development.
In many scenarios, they could build specific applications to solve concrete needs.
Applications that evolve with the business, that don't include hundreds of unnecessary features, and that can be modified without depending on a vendor's release calendar.
This introduces an even more radical possibility.
In certain contexts, software could become a disposable product.
Not disposable because of poor quality, but because its replacement cost would be low enough.
An application is built to solve a problem. When the problem changes, it's modified or replaced.
It isn't necessarily kept for fifteen years because the organization invested a fortune implementing it.
A license isn't necessarily renewed because migrating to another solution is too costly.
Software would stop being a hard-to-replace asset and become, in some cases, a tool that can be generated when it's needed.
It's an idea that deeply questions the sector's traditional economics.
But code was never the whole product
This is where it's worth avoiding an overly simplistic conclusion.
It would be tempting to announce that Microsoft, Adobe, Autodesk, and all the traditional vendors are on their way out.
But tech history is full of predictions about the disappearance of companies that, instead of dying, found new ways to compete.
Besides, the value of enterprise software has never been exclusively in its lines of code.
A banking system isn't worth only the features it implements.
It's also worth its reliability, its security controls, its audits, its integrations, its availability, and the contractual responsibility of those who operate it.
A company doesn't just buy an ERP system. It buys operational continuity, support, updates, specialized knowledge, and some guarantee that its critical processes will keep working.
An AI-generated alternative can reproduce visible features and still be far from offering those guarantees.
Building software can get cheaper.
Operating reliable, secure, sustainable software still has a cost.
And that distinction is fundamental.
Big tech companies have advantages that don't disappear just because it's now easier to generate code: integration ecosystems, user bases, recognized brands, certifications, infrastructure, and operational experience.
That's why we're probably not facing the end of commercial software.
We're facing a possible transformation of what customers will be willing to pay for.
When code stops being the product
If developing applications becomes progressively cheaper, value could shift to other parts of the chain.
Companies could stop paying mainly to access certain features and start paying more for integration, availability, security, regulatory compliance, specialized support, and business outcomes.
In other words, code could stop being the main differentiator of many products.
That doesn't mean all software will be free.
It means vendors will have to justify their price differently.
A vendor charging thousands of dollars a year for a relatively simple tool could run into trouble if its customers discover they can get equivalent results from cheaper alternatives.
On the other hand, a vendor that guarantees operational continuity, regulatory compliance, data protection, and flawless integration with business processes will keep offering something hard to replace.
Competition could shift from features to trust and quality of service.
And that would also change the work of technology professionals.
The most valuable knowledge might not be simply knowing how to implement a feature, but understanding what the business needs, how to design reliable systems, and how to use AI to build and maintain them efficiently.
Software engineering wouldn't disappear.
But its center of gravity could shift considerably.
Are we facing the next Spotify of software?
The comparison with Napster has limits.
The music industry went through a drastic reduction in the cost of copying and distributing its product.
The software industry is going through something different: a potential reduction in the cost of creating functionally equivalent products.
It's not exactly the same phenomenon.
And precisely because of that, it could have deeper consequences.
Napster didn't let just anyone produce a professional album in minutes. It let you distribute music someone had already created.
AI-assisted development tools, on the other hand, aim to democratize an important part of the production process itself.
That opens possibilities once reserved for organizations with considerable technical and financial resources.
It also raises new challenges.
If every company starts building its own applications, who will guarantee their security? Who will be responsible when they fail? How will they be maintained for years? How do we keep the apparent ease of producing software from generating a mountain of systems that are hard to manage?
AI can reduce the cost of writing code, but it could also multiply the amount of code we need to maintain.
Abundance doesn't automatically eliminate complexity.
And maybe that's where a new generation of tech companies appears.
Not necessarily companies dedicated to selling finished applications, but organizations specialized in generating, integrating, verifying, maintaining, and operating custom software through AI agents.
A model where the customer doesn't so much buy an application as the permanent ability to have the software they need.
A kind of on-demand solutions factory.
The beginning of a transformation
Crafting Apps could become a relevant project, or it could end up being one more among thousands of open-source initiatives.
It's still too early to know.
But its existence helps visualize a transformation that transcends any individual project.
Artificial intelligence is lowering some of the economic barriers that historically limited who could develop complex software.
And when those barriers start to fall, the business models built on them inevitably get exposed to new competitive pressures.
Music survived Napster.
Film survived digital distribution and streaming.
Commercial software can also survive artificial intelligence.
But surviving doesn't mean staying the same.
In a few years we may remember this stage as the moment we stopped thinking of applications as products we necessarily had to buy and started seeing them as solutions we could build.
And then, maybe, the question stops being how much a license costs.
The question will be how much it costs to solve the problem.
Napster taught us we didn't need to buy a record to listen to a song.
Artificial intelligence could teach us that we don't always need to buy a license to use an application.
And this time, we don't even need to copy the original product.
We can build another one.