The procurement conversation around enterprise Software as a Service (SaaS) is changing. It used to be “Can we afford to keep paying for this?” But it is increasingly becoming “Given what AI-assisted development tools can now do, should we still be paying for this at all?”
For New Zealand software companies selling to enterprise customers, that question matters. The assumptions underpinning many SaaS businesses were built in an era where internal replacement projects were expected to take years, required large engineering teams, and carried obvious execution risk. AI-assisted development has not eliminated those risks, but it has lowered some of the barriers to attempting an internal rebuild.
That does not mean enterprise customers will suddenly replace every SaaS platform they use. In many cases they will not. The hard part of enterprise software has rarely been producing a first version of the codebase. Operational maturity, integrations, compliance, reliability, workflow design, data quality, user adoption, and ongoing maintenance remain substantial advantages for established vendors.
But the threshold for an enterprise customer to seriously explore an internal alternative has undeniably shifted. That has implications for how software businesses think about intellectual property.
Copyright is narrower than many providers assume
Many SaaS providers instinctively look to copyright as the primary protection for software products. This is because in most countries, a computer program is protected by copyright automatically as soon as the code is created and recorded in a material form, without any need for registration. This automatic protection is commercially significant: it allows a software business to control use of its code, prevent unauthorised copying or piracy, and monetise a product through licensing, while recognising copyright as a valuable intangible asset of the business. Copyright absolutely matters, but it is important to understand what it protects and what it does not.
Copyright generally protects the expression of software code (i.e., as a “literary work” in almost all countries), not the underlying business functionality or outcome. An enterprise customer who independently recreates similar functionality without copying source code may avoid copyright infringement entirely. In other words, if the enterprise customer cleanroom designs his or her own version based on what your software does and needs to do, the enterprise customer does not generally infringe copyright. AI-assisted development tools potentially make that kind of independent implementation faster and cheaper than before.
That does not make copyright irrelevant. Direct copying, misuse of confidential material, derivative works, and unauthorised use of actual codebases remain serious copyright violations.
But copyright alone is rarely a complete moat for software businesses.
Patents may help but SaaS companies should be realistic
Patents are often discussed because they can, in principle, protect functional or technical inventions rather than merely the literal code implementing them. In the software context, however, the practical picture is more nuanced than many SaaS providers realise.
Software patents remain available in a number of major countries, particularly where claims are directed toward genuine technical improvements rather than abstract business processes implemented on generic computing infrastructure. In the United States, the ruling of the important Supreme Court case Alice v CLS Bank continues to make purely business-oriented software claims difficult to sustain, while more technically-grounded inventions may still have viable protection paths.
A claim directed to:
- “a method of using a computer to determine whether sufficient funds exist,”
- “a method of applying a mathematical model to predict behaviour,”
- “a method of managing customer support workflows” may struggle to be patent eligible in the United States.
Better claims directed to:
- “a specific technical architecture,”
- “a data processing technique,”
- “an AI model training approach,”
- “a latency reduction mechanism,”
- “a computer performance improvement” may stand a stronger chance
That said, software patents are not simple, cheap, or guaranteed solutions. Even well-drafted patents can face patent eligibility challenges, validity attacks (i.e., on whether it was properly granted by a patent office), and expensive enforcement processes.
Many highly successful SaaS businesses have been built with limited patent protection. Others have used patents effectively as part of broader strategic positioning with enterprise customers, particularly where there is genuine technical innovation at the infrastructure level. Even just having a “patent pending” status can help persuade others to not venture into your domain. And that is not an expensive proposition if done right.
For most software companies, patents should probably be viewed as one component of a broader defensibility strategy rather than a complete answer.
Trade secrets often matter more than providers realise
In practice, many software companies derive substantial defensibility from things that are never patented. AI training pipelines, operational processes, proprietary datasets, optimisation methods, deployment know-how, customer usage insights, and years of accumulated edge-case handling can all create meaningful competitive advantages that are difficult for competitors, or customers, to replicate quickly.
These present many opportunities for a multi-layered trade secret approach that can be extremely valuable, particularly where important aspects of the system are inter-related (you can’t do one aspect if you don’t know-how to do others) and are not readily observable from outside the platform. These aspects can be kept as a trade secret by restricting access, using well-drafted confidentiality and employment agreements, and implementing strong security and internal controls to prevent unauthorised disclosure.
Patents and trade secrets are not necessarily alternatives. In some software businesses, they layer effectively:
- Patents protecting selected technical innovations that can be seen by customers.
- Core operational knowledge and implementation detail remaining confidential.
The practical enforcement problem is real
One of the hardest scenarios for any software vendor is the internal customer rebuild. If a former enterprise customer creates an internal replacement system that is never sold externally, there may be very little public evidence available regarding how the system works or whether it infringes anything at all. That creates substantial practical difficulties for intellectual property enforcement.
Patent rights may theoretically extend to internal use in some countries, but pursuing those claims can be commercially and procedurally difficult:
- litigation is expensive
- customer rebuild information may not be discoverable in court
- customers may strongly resist intrusive investigations.
Thus, many software vendors will hesitate before initiating aggressive litigation against major enterprise customers.
Similarly, contractual mechanisms such as audit rights may help in some contexts. However, large enterprise customers often resist broad inspection provisions during procurement negotiations.
For many SaaS businesses, the practical value of IP protection is therefore partly deterrent and strategic rather than purely litigation-driven.
The real strategic question
The important question for software companies is probably not “Can AI help customers write competing code?” It clearly can. The more important question is “What remains difficult to replicate after the code itself becomes easier to produce?”
In reality, many successful SaaS companies have already spent years building forms of stickiness that are difficult to replicate quickly, even with modern AI-assisted development tools. For these platforms, the real moat was never just the codebase. It is deeply embedded workflows, customer-specific integrations, proprietary datasets, operational reliability, compliance frameworks, accumulated product refinement, user familiarity, support capability, security compliance, and the organisational knowledge required to maintain and evolve the system over time. AI may reduce the cost of writing software, but it does not automatically recreate years of operational maturity or customer trust.
Importantly, these defensibility layers are not fixed at company formation. SaaS businesses that recognise the shifts early can still strengthen their positions substantially over time. For some companies, the answer may genuinely involve patenting technical innovation. For others, it may be proprietary code and data protected as copyrights or trade secrets by attorney-approved non-disclosure agreements. Most likely, it will be some combination of all IP rights.
The key point is that software defensibility increasingly needs to be designed deliberately rather than assumed automatically. AI-assisted development does not eliminate the value of SaaS businesses, but it does force a more disciplined conversation about where their long-term moats sit. The likely outcome is not the disappearance of SaaS, but a more disciplined market where software businesses are forced to think more carefully about what genuinely makes them difficult to replace.
Our experts
Loren Tung is a qualified US patent attorney with over 20 years of IP experience, having worked for companies such as Samsung, IBM, Tencent and Rakuten.
Anton Blijlevens is a trans-Tasman patent attorney with over 32 years of experience advising New Zealand and Australian businesses on international IP strategy.
