Software Cloning – Is Paid Software Really Over?

Recent headlines of developers using AI tools to clone popular paid software – most notably Adobe – has sparked mixed reactions – from accusations of theft to welcoming free open-source tools.

But what does it mean for the tech companies who own the original software – or those who pay costly subscriptions to use it?

Matt Wood, Senior Director of Cyber Security Architecture, Operations, and AI Security at ABBYY, has investigated this growing concern, and gives his take on the risks and opportunities of AI-driven software development.

In 1996, Fatboy Slim built “Rockafeller Skank” from a guitar riff lifted from a 1960s record that few people had heard in decades. The Chemical Brothers looped breakbeats from forgotten funk records into stadium-filling anthems. The Prodigy sampled sirens, film dialogue, and obscure vinyl into something that sounded like nothing before it. Sampling did not just lower the cost of making music. It changed who got to make it, and it forced the entire industry to ask a harder question: what actually counts as original?

AI is forcing enterprises to ask the same question today, except the stakes are not chart positions. They are operational outcomes, compliance exposure, and the trust customers place in every decision a business makes.

Replication Is Cheap. Insight Is Not.

Large language models have made it remarkably easy to remix existing knowledge: draft a contract clause, summarize a report, generate a plausible-sounding answer from whatever text they can access. That capability has real value. But replication is not the same as discovery, and conflating the two is where many enterprise AI initiatives quietly stall.

The distinction matters because generative systems are very good at producing something that looks finished. They can imitate tone, structure, and confidence with extraordinary fluency. What they cannot do on their own is guarantee that what they produce is grounded in the right evidence, connected to the right process, or shaped by the right judgment.

That tension is not unique to business. It is visible in creative work too. A 2024 Science Advances study found that AI assistance improved the perceived quality and creativity of individual stories, especially for less creative writers, while also making the overall set of stories more similar to one another. In other words, the output improved at the individual level but converged at the collective level. More production did not necessarily mean more originality. It often meant more of the same ideas, expressed more smoothly.

That is the enterprise risk in miniature. When everyone has access to the same general-purpose tools, the competitive advantage does not come from generating more text, more summaries, or more recommendations. It comes from knowing what matters, what to trust, and what deserves action.

The Business Problem Is Not Generation. It Is Access to Reality.

Most businesses do not suffer from a shortage of language. They suffer from a shortage of accessible, reliable context.

What a company actually knows is usually scattered across contracts, invoices, onboarding forms, claims, emails, reports, scanned PDFs, and the workflows that shuttle those documents from one team to another. A payment exception in finance, a compliance issue in onboarding, or a service delay in operations often begins life as something half-structured, inconsistently formatted, and disconnected from the systems leaders rely on to make decisions.

That is why so many enterprise AI projects feel impressive in demos and can fail in production. The model may be capable, but the context is partial. It can only reason over what it can reach, and in many organizations that is a narrow slice of the real picture.

This is not just a theoretical weakness. It affects the quality of decisions. If a model summarizes a contract without access to the relevant amendments, it may sound clear while being incomplete. If it analyzes accounts payable issues without visibility into invoice exceptions across formats and suppliers, it may produce recommendations that are neat but not useful. If it helps with customer onboarding while key information remains trapped in scanned forms or fragmented workflows, it can accelerate the wrong conclusion.

The result is not insight. It is confident-sounding guesswork.

Why Imitation Scales Faster Than Understanding

This is where the music sampling analogy becomes more than a metaphor.

Sampling made music production faster, cheaper, and more accessible. But the producers who lasted were not the ones who merely copied fragments. They were the ones with the taste and discipline to transform them. Anyone could lift a beat; not everyone could build a new sound around it.

AI is creating a similar divide in software and knowledge work. The barrier to imitation has fallen dramatically. It is easier than ever to clone interfaces, reproduce outputs, mimic workflows, and approximate expertise. Entire categories of software increasingly risk becoming reproducible at the surface level.

That does not mean paid software is over, as pronounced by developer Brandon Thomas who has created free alternatives for Adobe. It means the value of software is shifting.

When imitation becomes cheap, value moves elsewhere: to trust, integration, judgment, and the ability to produce outcomes that hold up under real-world conditions. The question is no longer whether an AI system can produce an answer that resembles expert work. The question is whether it can do so reliably, in context, and with enough transparency that people will trust it when the stakes are high.

A cloned output is easy to produce. A dependable outcome is not.

The Hidden Premium Is Expertise

This is why expertise becomes more important, not less, in an age of AI cloning.

For years, many digital tools derived their value from making specialized capabilities scarce. If software could encode a process, standardize a task, or simplify a previously difficult workflow, customers would pay for access to that packaged expertise. Generative AI unsettles that model because it can now imitate many visible features of expertise at low cost.

But imitation only captures the outer layer.

Real expertise is knowing which exception matters, which ambiguity cannot be ignored, which source is authoritative, and when a system’s answer should be challenged rather than accepted. It is the difference between extracting text from a document and understanding whether that text is sufficient to support a business decision. It is the difference between summarizing policy and recognizing the operational consequences of getting that summary wrong.

This is also where many discussions of AI and originality go astray. Originality in business is rarely about novelty for its own sake. It is about seeing something others miss: a risk hidden in process noise, a pattern buried in fragmented information, an opportunity obscured by disconnected systems. That kind of originality depends less on eloquent generation than on accurate interpretation.

Documents Are Not the Whole Story, but They Are Often Where the Story Starts

Documents matter here because they remain one of the most common places where enterprise reality is stored but not fully usable.

That does not make “Document AI” the hero of the story. It simply makes document understanding one part of the larger discipline of turning operational information into something AI systems can work with responsibly. If organizations want better outcomes from generative models, they need better access to the evidence those models rely on. In many cases, that means extracting, validating, and structuring information that would otherwise remain trapped in forms, files, and fragmented handoffs.

Used well, that kind of capability is not about promoting one product category over another. It is about reducing the distance between what the business knows and what its AI systems can actually see.

That distinction matters. The strongest enterprise AI strategies are not built on generation alone. They are built on the combination of usable data, process awareness, and human oversight. Recent enterprise research from MIT’s Project NANDA, for example, argues that the biggest barriers to AI adoption are often not model quality in isolation but poor workflow integration and a lack of trust in how tools behave in real use.

That should sound familiar. Businesses do not win by producing the most synthetic output. They win by making better decisions, faster and more consistently, from information they can defend.

The Next Premium Is Judgment

Sampling did not devalue music. It changed what originality meant.

Once the raw materials became easier to reuse, the premium shifted to selection, arrangement, and intent. Judgment became more valuable than access. The same thing is happening now in enterprise AI. As software cloning gets easier, the durable advantage will not come from having a model that can imitate competent work. It will come from knowing how to connect AI to the right context, constrain it appropriately, and apply it where human creativity and domain judgment still matter most.

That is the part of the conversation many “AI replaces software” arguments miss. If every tool can be copied at the interface level, then the real moat is not the interface. It is the depth of understanding behind it: the workflow logic, the exception handling, the trusted data, the institutional knowledge, and the human expertise required to turn outputs into decisions.

So, is paid software really over? Only if we believed software’s value was ever just the visible layer.

For the commodity features of software, pressure will intensify. Some products will be cloned. Some categories will compress. Some functionality that once commanded a premium will become table stakes. But the tools and companies that endure will be the ones that help organizations move from imitation to understanding.

That is why AI’s sampling moment matters. It reminds us that lowering the cost of reproduction does not eliminate the need for originality. It simply changes where originality lives.

In music, it moved from the performance of raw material to the craft of transformation.

In enterprise AI, it is moving from the generation of plausible outputs to the disciplined use of trustworthy information, expert judgment, and creative problem-solving.

Replication is cheap. Insight is not. And that is why the next phase of AI will not belong to the systems that merely sound intelligent. It will belong to the organizations that know how to make what they already know usable enough to discover something genuinely new.