“Base-case: Agentic AI reshapes, but does not annihilate, the software stack. Front-end interfaces and some consulting-heavy workflows will be commoditised, pushing headline growth multiples and gross margins lower. However, the data models, compliance layers and domain-specific logic housed in incumbent SaaS back-ends remain sticky and become the rails upon which agents run. The market’s “AI eats software” panic therefore overshoots: value migrates from UX and billable hours toward API breadth, latency-optimised infrastructure, and usage-based pricing. Legacy vendors that rapidly expose clean, usage-metered endpoints and embed first-party agents can re-accelerate; those that cling to seat licences or depend on captive professional-services revenue will keep de-rating. The structural friction across pillars resolves into a bar-bell outcome—compression, not eradication, of traditional software economics, coupled with a new layer of demand from agent orchestration.”