In recent months, a catchy narrative has been making the rounds: SaaS is dead. Since AI has made writing code so trivial, the argument goes, every company will soon simply build its own software. So why keep paying for expensive subscriptions? <\/p>\n\n
That is a fundamental misconception. And above all, it shows just how little understanding there is of what actually defines a commercial software product. <\/p>\n\n
(Reading time: 4 minutes)<\/em><\/p>\n\n SaaS is a mystery. It is nothing more than a software product delivered in a way that is suited to the cloud era. A good software product typically solves one or more specific problems for the user. Most of the time, it\u2019s about efficiency. <\/p>\n\n This requires three things: First, a deep\u2014and often painful\u2014understanding of the problem. Second, code. And third, distribution. <\/p>\n\n The first point in particular\u2014a deeper understanding of the problem to be solved\u2014is almost systematically underestimated. This is because successful off-the-shelf software that has been on the market for years does not merely \u201cconvey\u201d code; rather, it is, so to speak, the distillation of decades of \u201cbest practices.\u201d When a customer purchases software, they are always also buying the domain knowledge embedded in the product. It goes without saying that this does not initially provide an unfair competitive advantage, but merely brings the company up to the industry standard. For most companies, however, that is already a great deal. <\/p>\n\n Yes, the AI boom is making it dramatically easier to write code. Today, anyone without programming experience can \u201cprompt\u201d their way to software components. That\u2019s great, but the conclusion that companies are now building their own software across the board is far-fetched. The reasons for this are of a business nature. <\/p>\n\n First:<\/em> Building software has always been easy\u2014at least for those who know how to do it. In all the years I\u2019ve been building software with my teams, the code has never really been the deciding factor in whether a company succeeds or fails. If AI changes anything, it\u2019s primarily by increasing efficiency in software development. <\/p>\n\n Second:<\/em> Costs. Anyone who bases a make-or-buy decision solely on initial development costs is making a miscalculation. Even though AI can now generate code almost autonomously in just a few hours, the actual writing of the software typically accounts for no more than 20 percent of the total costs over the entire software lifecycle. <\/p>\n\n The remaining 80 percent goes toward operations, maintenance, bug fixes, adaptation to new operating systems, security updates, and scaling. In the best-case scenario, a SaaS provider spreads these massive fixed costs across thousands of customers. A company with an in-house solution bears 100 percent of this risk on its own. <\/p>\n\n Third:<\/em> Software \u201crusts.\u201d Code is not a static entity. External APIs are constantly changing, new security vulnerabilities emerge, and data structures grow. Customers\u2014both internal and external\u2014literally \u201crain down\u201d requirements. Who will maintain the AI-generated in-house software in two years when the highly motivated employee who built the AI-powered solution leaves the company? You\u2019re creating a dangerous legacy black box. <\/p>\n\n A logistics company must excel at transporting goods. An agency must design creative campaigns. It makes no sense to tie up valuable resources to build a CRM or invoicing tool that\u2019s available on the market for just a few euros per month. In-house development only makes sense when it directly supports the company\u2019s core business and generates a genuine, unique competitive advantage\u2014and when the in-house resources are available to operate and maintain such software over the long term. <\/p>\n\n Does that mean established SaaS providers can now sit back and relax? I don’t think so at all. On the contrary, I think many of them are in serious danger. <\/p>\n\nNot SaaS, but a software product<\/strong><\/h2>\n\n
Understanding the Problem. Code. Distribution <\/strong><\/h2>\n\n
Code was never \u201cthe hard part\u201d<\/strong><\/h2>\n\n
Focus on Strengths Within the Company<\/strong><\/h2>\n\n
From Tool to Digital Employee<\/strong><\/h2>\n\n