The buzz around AI-powered software creation is more than just hype. Following a survey of Stack Overflow from June 2025 among more than 49,000 developers, a steadily growing 65% of them use AI tools weekly or even daily. Yet, this gold rush isn’t without its pitfalls. AI coding assistants still struggle when tasks become too complex, as only 30% think they are good at this.
They struggle with the nuances that are needed for large-scale, secure solutions. Which raises critical questions for the tech industry: where is this technology heading? How will it reshape the business models of Software-as-a-Service (SaaS) companies – whose core capability, in effect, is the ideation, creation and operation of good software?
While today’s AI coding assistants are largely adding value by accelerating junior-level tasks, the immense investments that are pouring into the field suggests that this is just the beginning. Such funding – and the sheer demand from developers – drive leading tech companies, start-ups and open-source communities to address the numerous challenges these assistants show. We can realistically expect incremental innovations to push this technology to a point where it can accelerate even senior-level coding tasks.
While it is hard to predict how long it will take – once we have reached it, this evolution will fundamentally rewrite the rules of the SaaS industry. It will profoundly change what makes a software company competitive. For years, a unique set of features proved a defensible moat. Think of how Dropbox pioneered the automatic file synchronization, or how Airtable blended the usability of a spreadsheet with the power of a database.
These were advantages, realized through code that offered unique features that were hard to copy. But as AI will make the replication of such features much faster and cheaper, this competitive moat will melt.
We will find that the costs of creating this logic layer in software will become cheaper. Further, the foundational hardware layer that powers the AI assistants is increasingly dominated by a few large cloud players. Ultimately, this will shift the true battle ground for differentiation “up the stack” towards their users – SaaS vendors will need to own and master the user experience in order to survive.
The paradox of automation and the hidden cost of speed
However, this new reality creates a paradox. The intense pressure to innovate and deliver solutions at an ever increasing speed comes with a hidden cost. High-velocity software development, when paired with imperfect AI tools that lack awareness of the architecture in the big picture, can create a massive maintenance mortgage. In consequence, technical debts accumulate and harm agility and security of solutions in the long-term.
This cannot be counter-acted by mere software development. The roles of software engineers – professionals who architect optimal software structures – will become more critical than ever. Creating code that works as fast as possible through AI will not be enough – the single components need to work together.
But the changes reach deeper as the very structure of software will set out to transform. With the advent of agentic AI architectures, software will become less of a single large block of code and more of a composable collection of capabilities.
In essence, AI agents will be the brains which translate users’ intents into concrete tasks. They will have access to numerous specialized capabilities, each one crafted to optimally realize a specific request, and orchestrate how these capabilities will be used. How these capabilities will act together might need to change from customer to customer. Yet, the AI agent’s “brain” will be able to take care of this.
Instead of hard-wiring specific workflows into software that limit what users are able to do, agents maximize users’ flexibility. And therefore set the stage to build hyper-customized solutions. Yet, the simplicity and low cost with which software can be extended with new features will also place immense pressure on traditional SaaS vendors. How can they justify their pricing and keep on proving their unique value?
How to differentiate when code becomes ubiquitous
Consequently, they will require new competitive moats that go beyond the features built into software alone. Sustainable strategies will stem from means of defense that are intrinsically hard to replicate.
As a first example, network effects that are based on engaged users and ecosystems of developers who are hooked in through their own contributions can build powerful approaches. Proprietary data of niche domains, processes or knowledge which no competitor possesses in comparable quality and volume can be used to train best-in-class AI models for certain verticals. The sheer velocity of innovation itself becomes a moat in this fast-paced environment when a company consistently delivers faster than competitors. Further, the deep integration of software with physical assets, such as IoT devices, introduces components that cannot be created through AI assistants.
This ensures the competitive dynamic isn’t just about price. A competitor might replicate a feature, but they cannot replicate an engaged ecosystem or a decade of proprietary data. The choice for a customer, therefore, becomes a strategic trade-off, not a simple cost comparison.
Ultimately, however, the most crucial differentiator will be the human aspect. Vendors can own and master the user experience only when connecting with and understanding their customers through human skills that go beyond purely technical know-how.
This poses a two-fold amplification of the human factor that impacts both ends of the software lifecycle. Not only are deep bonds to users amplified. During creation, the software engineer’s role is elevated to that of a strategic designer who shapes solutions that meet nuanced customer needs.
New eco-systems shape new pricing models
From a commercial viewpoint, rigid per-seat pricing models of SaaS solutions will become obsolete as AI makes it easier to amplify human productivity. After all, when AI’s promises to significantly boost productivity of single employees turns into reality, charging by the impact achieved instead of the number of users will become the relevant benchmark. This gives way to models where a vendor’s fees are directly tied to the value created for customers.
These aren’t just theories – pioneers are already starting to make this shift. Zendesk, for instance, now charges for its AI agents only when they successfully resolve a customer issue without human help, and companies like Riskified and Chargeflow tie their fees directly to preventing or recovering fraudulent chargebacks for their clients.
So, where is this technology truly heading? While the initial excitement focuses on productivity gains and the automation of code, the real destination is far more transformative. The ultimate impact of AI coding assistants isn’t the replacement of human software engineers, but their amplification. While developers who are merely creating code will struggle in the future, the role of humans in the software lifecycle gets transformed.
This transformation is also forcing a fundamental shift of how we measure the value of software, as the most defensible asset is no longer code itself. Instead, the humans relationships built around it will be crucial.
This inevitably leads us to deeper connections between vendors and customers. When customers’ success will be directly tied to the earnings of SaaS vendors, we will face interconnected ecosystems where successes of multiple companies are intertwined stronger than we know today. In a more human-centric software economy, we will win and lose together.
