The emergence of LLMs, vibe coding, AI agents and a mature ecosystem of open-source and self-hosting tools has altered the core build-vs-buy equation. Building software is no longer an exclusive privilege reserved for large corporations or category leaders. It is becoming a practical option for a much wider business base.

At this point, a company with a modest technical team, or even with a handful of hybrid non-programmer profiles, can deploy functional tools on its own infrastructure at a cost and within a time frame that would have seemed unrealistic just a couple of years ago. That alone changes the balance of power between software vendors and customers.

Another reason this shift is hard to ignore is the reaction of financial markets. Valuation multiples across many major SaaS companies have fallen sharply, pointing to real concern about the long-term value that iconic firms such as HubSpot may be able to sustain. Something is shifting beneath the surface, and some of the main tectonic plates of the technology sector are clearly moving. Those movements create friction. The key question is whether this is a structural trend or merely a cyclical adjustment.

Subscription-led digitalization and its three dependencies

The digitalization path followed by most small and mid-sized companies has been remarkably similar for years: successive layers of SaaS piled one on top of another. Each new subscription solved a specific problem quickly and without requiring internal technical resources. For a long time, backing those platforms was the most sensible option, and that choice was rarely debated because it genuinely made sense.

The model was rational: entry barriers were low, updates were included, and infrastructure responsibility sat with the provider. In most cases, the best option was to adopt proven tools that covered a market niche efficiently. Incentives were strong, and it is precisely that incentive structure that is now starting to wobble. Does it still make sense, for example, to rely on WIX as the core of a web architecture? It is increasingly difficult to answer yes without qualifications.

Returning to the broader issue, a decade and a half into a long SaaS boom cycle, the digital landscape of most companies is functional yet strategically fragile and far from optimal. Firms have accumulated, on average, between 24 and 62 SaaS applications when they employ between 50 and 500 people. Larger organizations exceed 177 applications. In many cases, finance departments discover that almost half of purchased seats have not been used in the last thirty days. There is nothing revolutionary in saying this, but the numbers are too forceful to dismiss.

At the core, three structural dependencies have become increasingly visible over time.

• Financial dependency: licenses are often expensive and their cost tends to rise faster than inflation. This also applies to AI-related tooling. Every company now needs to examine the concrete value each tool delivers and whether that value truly justifies its price.
• Technical dependency: functionalities and processes are usually shaped by the logic predefined by vendors. Deep customization often requires costly integrations or additional development. As a result, companies only partially benefit from standardized software while still paying to adapt it.
• Strategic dependency: these technologies generate significant switching costs. Once an organization becomes accustomed to a tool, replacing it can be traumatic and difficult. At a certain scale, adopting a specific technology stack creates strong incentives to keep solutions that may be functional but are no longer optimal.

CRM as an illustrative symptom

CRM is probably the clearest use case for understanding the tension now running through the software industry. The SaaS CRM market would currently stand at a global size of around $66 billion, with a growth rate close to 20% CAGR. Platforms such as HubSpot ended 2025 with nearly 289,000 paying customers.

At the same time, usage and satisfaction data point to underlying frustration. A relevant share of mid-market CRM users only relies on a fraction of the available functionality, often because of poor training, a mismatch between the product and the actual sales process, or simply because too many features add complexity without adding value. Most companies do not need everything they end up paying for, and that functional overload makes day-to-day work harder for the people inside the platform.

The accumulated cost of a mid-tier SaaS CRM for a small team can reach between €15,000 and €50,000 over a three-year period, and it can go much higher once users, modules and third-party integrations are added. For many organizations, that is a substantial amount of money, so it is entirely reasonable to question whether it is being spent wisely.

Meanwhile, the market for open-source and self-hosted alternatives has matured considerably. Options such as SuiteCRM, EspoCRM, Twenty or Vtiger offer functional value comparable to that of paid platforms. With zero license cost and a total cost of ownership that can be between 80% and 90% lower for teams above ten users, these tools may not replicate the full value proposition of Salesforce or HubSpot, but they can serve a very large share of organizations. That is where the tension lies. HubSpot is a formidable tool and covers the needs of most companies. And where HubSpot does not reach, Salesforce often does. The problem is that both integrate and charge for features that most of those companies will never use. Do those organizations have the capability to implement, adapt and sustain lighter tools over time? That is the key question.

The short answer is that not many companies can continue to justify the cost of a traditional SaaS CRM, and that same logic is spreading into many other corners of the technology market. The pace of change is fast.

LLMs, vibe coding and the generalization of development capabilities

From bottleneck to multiplier

For decades, the main obstacle to building custom software has been the same: the scarcity of programmers and the high cost of hiring them. Putting other headaches aside, development pricing and limited access to reasonably priced technical talent were among the strongest barriers against building internal tools.

Building an internal CRM, a workflow automation tool or a dashboard on top of proprietary data usually required a stable engineering team, a long development cycle and a budget that very few mid-sized companies could justify. SaaS thrived because it filled a real market gap, delivered value and often produced results from day one.

Between 2024 and 2026, the ability of LLMs to generate functional code has improved dramatically. Available data suggest that around 41% of the code written in 2025 already originated in AI suggestions, that 84% of developers use or plan to use AI tools in their workflow, and that 51% of professionals use them daily. At companies such as Microsoft and Google, between 20% and 30% of code in internal repositories is already AI-generated. In startup environments, the share can rise as high as 95%.

The effect on cost and development time is measurable. What recently required ten months of work and a budget above $100,000 can now be built in three months for between $20,000 and $40,000, thanks to tools such as Claude Code, Cursor and GitHub Copilot. That radically alters the economics of software creation.

Vibe coding: a paradigm shift that goes beyond developers

The term vibe coding, first popularized by Andrej Karpathy, describes a development practice in which users describe what they want in natural language and AI produces a highly advanced code proposal. What once looked like an experimental concept is now already mainstream enough to let many people with limited programming skills build useful tools for their companies.

Its impact is at least twofold.

• On the one hand, available developers become substantially more productive. Studies show productivity gains of between 25% and 39% in coding tasks and between 50% and 70% in boilerplate, testing and documentation.
• On the other hand, a structural change takes place: teams without strong technical muscle can now orchestrate the creation of functional tools without relying entirely on engineering departments. Anyone who has run even a few experiments with these systems understands the shift.

The gap between a functional prototype and a production-grade tool with security, permissions and real data integration still exists and still matters. Even so, vibe coding democratizes the starting point and drastically reduces experimentation time. It does not eliminate the need for technical judgment in deployment and maintenance, but it clearly changes the overall workflow. That is exactly why this is such fertile ground for digital integration firms and why part of the technology sector will try to reinvent itself around this disruption.

GitHub and far beyond

Media attention has recently focused heavily on GitHub, but the phenomenon is much broader. GitHub and similar platforms have moved from being code repositories to becoming warehouses of reusable capabilities. There are now millions of mature open-source projects, coding agents, deployment frameworks and builders that make it possible to assemble internal tools without starting from scratch. Anyone working anywhere near digital operations feels that impact daily.

To take one example, the AI software agent market was valued at $10.4 billion in 2025 and is projected to reach $149.6 billion by 2034. Internal-tools platforms such as Retool, WeWeb or Glide make it possible to build interfaces, workflows and integrations on top of proprietary APIs and databases, with self-hosting options and full ownership of data. The open-source CRM segment is growing at a 13% CAGR and already offers mature alternatives for most mid-market use cases.

The result is an ecosystem in which the marginal cost of owning software has declined in a sustained way, and in which the main barrier to having proprietary code is no longer just scarce technical talent. It is the ability to structure the problem correctly, design a reasonable architecture and guarantee long-term maintenance. All three are areas where external integrators can play a critical and economically justified role.

Financial markets react

At the beginning of 2026, the software sector went through a sharp market shock in terms of capitalization. Put simply, consensus about the value of SaaS companies changed, and that shift reflected a belief that SaaS was becoming less attractive as a business model. Horizontal platforms such as Salesforce, HubSpot and ServiceNow experienced stock declines of between 30% and 50%, while specialized infrastructure players such as Cloudflare moved up. In markets, accidents happen, but broad repricing rarely comes without reason.

Among the factors associated with this moment, the Klarna case stands out. Its CEO, Sebastian Siemiatkowski, announced that the company would phase out tools such as Salesforce and Workday as part of a broader consolidation effort involving more than 1,200 SaaS tools into a unified internal data framework supported by AI. The case received major media attention, but it is likely only the visible tip of a larger shift.

Retool’s survey captures the real dimension of what is happening. Thirty-five percent of participating organizations had already replaced the functionality of at least one SaaS tool with an internally built solution, and 78% planned to build more internal tools during 2026. Is this the beginning of a much larger wave? It cannot be ruled out. The categories under the greatest substitution pressure include workflow automation, internal administration tools, BI and dashboards, CRM and sales tools, project management and customer support. The fact that pressure is visible across all categories is significant and suggests the emergence of a broad trend.

There is a growing segment of organizations with enough technical capability to actively replace SaaS tools with proprietary solutions in specific categories. There is also a broader segment that still prefers to buy, but does so with more realistic expectations and stricter purchasing criteria. Financial markets have already noticed this, and future price action will likely offer useful clues about the direction of major SaaS businesses beyond corporate messaging.

Who may benefit and who has a lot to lose

The shift in the build-vs-buy frontier will not automatically benefit every company, and that should be clear from the outset. Building proprietary software means taking on design, maintenance, security and evolution costs that many organizations systematically underestimate. Building without governance often creates disconnected internal tools that are hard to audit and difficult to maintain over time. Riding this wave has a cost. It may be lower than licensing software, but it is not zero.

This is precisely where companies capable of positioning themselves as digital integrators will find their opportunity. They can help organizations decide, with objective criteria, where it makes sense to build, when it makes sense to buy, and how to combine selected solutions into an optimized and context-specific tool landscape. Digital integration is not, and certainly will not remain, a marginal service. It is becoming a first-order strategic lever.

The fact that the technology landscape is moving so quickly also creates a divide between companies that can keep up and those that cannot dedicate time to doing so. Keeping up is not just a matter of consuming news. It has much more to do with forming judgment and translating that judgment into an actionable vision of the future.

The decision to build software or buy it has not changed in nature, but it has changed in scale. Where the question barely existed for companies without large technical teams, it is now legitimate for a much wider range of organizations. Build or buy? The answer increasingly depends on the specific case.