AI does not signal the end of large SaaS, but changes its use. Their future involves agentic, agile and tailor-made AI overlays, complementary to existing reference systems.
Will AI kill legacy SaaS such as Salesforce.com, SAP, Outlook or ServiceNow? Will a new generation of agentic SaaS specialized by function (horizontal) or by industry (vertical) emerge?
The plate tectonics at play are complex but the most likely medium-term scenario is the survival of traditional players, more weakened than strengthened by AI, and the return to modular and agile tailor-made solutions (“customs”) in addition to historical “systems of record”.
“Big” legacy software has been expensive and complex to deploy, providing a backbone of systems of record and reference. In this sense, they are quite impregnable, in the medium term.
However, their ability to adopt AI is limited. Certainly, they deploy great resources but their technical architecture is fundamentally “from another era” and their prism is limited, whatever they say, to the limits of the transactions they support, when the issue of adopting AI in business is, by definition, transversal. In addition, these historical software programs have been enhanced with various extensions over time, often via more or less integrated acquisitions (for example, Ariba for SAP, Datorama or Slack for Salesforce.com). These peripheral solutions are the most “attackable” by AI which can provide an elegant solution to supplier interaction, marketing performance analysis or optimization of a collaborative workflow.
In the longer term, can legacy SaaS maintain their market share and competitive advantage?
AI is lowering the barriers to entry into the industry and the threat will grow as the technology improves. You can already code a simple but complete CRM software in a weekend with the help of Claude or Gemini. Specialized software is at greatest risk. You can develop a website for a few euros with Lovable, the tools for generating a totally realistic video from a prompt and a simple photo of the protagonists are legion, soon commercial sites or marketplaces will be able to be generated with the same simplicity using new tools. Gencore, New Gen, Qeen AI, Graas, Sellful, Appscrip AI, Perplexity commerce are positioning themselves, Google is trying to redefine e-commerce in prompt-to-store mode to disintermediate sites: what future for Adobe, Figma, Shopify or Mirakl? Certainly, the millions of lines of code in an SAP or a Salesforce.com accumulate valuable business knowledge, but how long will this barrier hold against competition augmented by increasingly powerful AI?
In parallel with this dynamic, specialized peripheral business applications are currently emerging in large numbers: contract management, drafting of secure content for regulated industries, management of insurance claims, etc. Some will perhaps impose themselves but the majority will be swept away by the law of the market. Their added value in relation to the creation of tailor-made solutions perfectly adapted to the needs of the company, based on basic, modular and scalable AI building blocks, is questionable. Their “lock-in” model by payment of subscriptions increasing with volumes and periodically revalued upwards is unattractive. It is not guaranteed that these players have the capacity or willingness to adopt the technological innovations of the fundamental models in order to improve their performance.
Finally, these new software quickly tend to put the “body weight” on business functionalities and user interfaces, thus finding themselves duplicating existing transactional (“reference”) systems, without really being able to replace them. Conversely, the winning model is to create in the short term, above large reference systems, a “tailor-made” overlay directly calling on utility bricks in a transparent and scalable “pipeline” of technologies.
This agentic overlay is “chemically pure”: it automates or augments the activities of human users in traditional systems in a way that is totally centered on the contributions of AI. For example, instead of purchasing a new AI-native SaaS which schedules team interventions by duplicating the functionalities of the existing planning system (which is already integrated with the HR system and customer invoicing), a service company would be wiser to develop a lightweight optimization overlay, according to its own objectives and constraints, which injects an optimized planning into the existing system without duplicating it, without creating a new layer in the already complex application millefeuille, without requiring hours of user training in a new interface.
AI is reshuffling the cards in the software market, lowering the barriers to entry for challengers of current heavyweights, but above all providing a huge opportunity for companies to regain control. These can increase their negotiating power vis-à-vis historical technology suppliers but also create real competitive advantages thanks to agile, modular, scalable and tailor-made solutions which complement them in an agentic use dimension.
