AI is shifting the value from software applications to software orchestration
Is that a Meteor approaching?
Since the start of the year SAP and Salesforce stocks have lost significantly - the trend of valuation losses has engulfed the entire SAAS industry since early 2025. While several factors contribute to declining SaaS valuations, investors and stakeholders appear increasingly concerned about one question:
What happens when AI becomes the interface layer between users and software applications?
Resilience Required
The SaaS business model typically sells software products and charges customers per user. With AI replacing human workers it eats into the profits of legacy products, as AI agents eliminate ‘seats’. This is also true in case established SaaS companies offer their own AI tools. The costs of AI tools however are driven by usage, which leads to a tricky question: How to adapt the SaaS business model to AI?
The Four Horsemen of the Apocalypse
Horeseman #1: AI reduces the time and the cost of code generation
AI’s recent coding successes have been remarkable. In 2019 it took OpenAI 168 hours and 32 state-of-the-art chips to develop its GPT-2 chatbot. In 2026 Andrej Karpathy of OpenAI built a similar chatbot with the help of AI in just under three hours and eight off-the-shelf GPUs.
Horseman #2: SaaS sells applications, AI sells orchestration
Even more telling is SaaS companies’ failure to break out of their specializations. Even a best in breed strategy, as is common among SAP’s or Salesforce’s clients, leaves them at the mercy of SaaS companies’ support for integration and adaptation. Big AI labs on the other hand operate on a higher level and move horizontally across functions and industries. On top of SaaS products they weave a mesh of programs by connecting them with plug-ins that can be controlled by a single chat-based interface. Today a user can ask an AI assistant to gather information from CRM, ERP, email and calendar systems through connectors and present a unified answer. Traditionally, this required navigating several applications individually.
Horseman #3: AI-native startups
Another factor for the ongoing slump of the SaaS industry is the rise of numerous AI startups that deliver customer-service specifically designed to clients’ needs. One such startup, Sierra, was co-founded by Bret Taylor, the chairman of OpenAI and former co-chief of Salesforce. But there are also many other AI-startups that reach a valuation of 1 billion Dollars in two years.
Horseman #4: DIY software creation
The fourth threat to the SaaS business model is that customers can use AI to build DIY software or even their custom software systems. The question to either ‘build or buy’ is being reevaluated across various industries, ever since progress has moved well beyond ‘vibe coding’ simple tools in order to automate tedious tasks.
Why orchestration matters…
Looking ahead none of this seems to be an imminent threat that might bring down the SaaS industry. SAP and Salesforce are still leading firms with armies of sales personnel and consultants. SaaS incumbents possess significant advantages. Deep customer relationships, trusted ecosystems, and decades of process knowledge may prove harder to disrupt than many AI enthusiasts assume. If SaaS companies manage to adapt their business models by integrating the capabilities of AI, the industry might soon boom again.
Yet this might require a shift in paradigm: Users don’t care about applications, they care about results. A browser, an operating system, a search engine all serve similar purposes: to reduce complexity in the eyes of the user and thus enabling her to achieve the desired results.
The big question will be, how rapidly SaaS companies can reap the benefits of AI and offset losses, before customers buy fewer licences - a typical cannibalization problem. And herein lies the irony: A well known slogan in the software development industry is that “Software eats the world”. Who would have thought that software once would eat itself?
Comments