AI is set to swallow half of SaaS. Arthur Mensch, CEO of Mistral AI, drove home this turning point in a CNBC interview at the India AI Impact Summit in New Delhi, saying more than half of what enterprises buy as SaaS could shift toward AI-built applications, thanks to faster software creation and tighter coupling between data and workflows.
This comes as investors reprice software, betting that AI will capture value from SaaS, especially after recent news from Anthropic on Claude Cowork and automation.
Why this is more than a headline
Enterprise SaaS has historically scaled by standardizing workflows and selling them broadly. AI shifts the center of gravity from standardized user interfaces toward dynamic, data-connected execution. Mensch’s point is that once a company has the infrastructure to securely connect internal data to AI systems, teams can build custom workflow applications in days, for areas like procurement and supply chain.
This changes the economics of software in two ways.
First, it lowers the cost and time required to create workflow tools. If teams can assemble fit-for-purpose apps quickly, the premium paid for generic workflow SaaS compresses.
Second, it enhances the strategic value of ready, well-governed internal data. Success will depend on strong data management for AI.
Two software categories diverge
Mensch distinguishes that systems of record, platforms storing core data, remain essential. AI relies on these rather than replacing them.
Rubrik CEO Bipul Sinha echoed a similar split in comments to CNBC, arguing that workflow software faces greater disruption, while data and infrastructure software that enables AI can benefit from the transition.
So the likely near-term pattern looks like this:
- Workflow tools face pressure as AI becomes the new application layer.
- Systems of record and data infrastructure gain importance as sources of truth.
What “AI replaces SaaS” looks like inside a company
In practice, replacement usually means unbundling.
Classic SaaS bundles UI, logic, workflow, analytics, and integrations. AI can unbundle them: the conversational interface is the UI.
- Business logic becomes prompts, tools, and policies tied to systems of record.
- Integrations shift to API-level building blocks.
- Governance becomes a first-class requirement because AI actions need permissions, logging, and evaluation.
This is why transitions start in workflows with clear, measurable outcomes.
India as the next proving ground
Mistral’s comments also came with a geographic signal. Mensch told CNBC that the company plans to open its first office in India this year and to pursue both public- and private-sector customers. He also pointed to India’s interest in local model execution and domestic data storage, as well as the need to support multiple languages.
AI players are using Delhi's summits to deepen India strategies and stress scale.
The investor's question underneath the story
Investors are trying to answer a simple question: where does pricing power move?
If AI shifts value away from packaged workflow SaaS, multiples compress for categories that depend on standardized workflows and seat-based pricing. At the same time, categories tied to data infrastructure, security, and governance can expand as AI becomes more important and receives higher spending priority.
The headline “50%” is provocative. The deeper point is that enterprise software spend is being resegmented into:
- Data and control layers that make AI safe and reliable
- Execution layers where AI can assemble workflow apps rapidly
- Legacy SaaS categories like ticketing systems (Zendesk), expense management (Concur), and CRM tools (Salesforce) must evolve into AI platforms or risk commoditization, while data warehousing (Snowflake) and identity management (Okta) are more insulated.