The 'AI disruption premium' is turning into a disruption discount
Public markets are acting like they've finally internalized something product teams have been muttering about for months: if LLMs can execute workflows end-to-end, a chunk of today's SaaS value may be packaging, not defensibility.
A recent sell-off was widely linked to investor fears that new AI automation capabilities could compress demand for parts of the traditional software stackespecially where features look like they can be replicated by an agent sitting on top of a model.
Why this matters beyond a single red day
- The market isn't just rotating between 'AI winners' and 'AI losers'; it's repricing the idea of what a 'platform' even is when models can consume tools, APIs, and data on demand.
- If you're a software exec, the uncomfortable question is: are you building a product customers love, or a feature set that a model can reconstitute from prompts + integrations?
What to watch next
- Expect faster product cycles where vendors ship agent-friendly primitives (auditing, policy controls, sandboxed execution) to prove they're more than a UI layer.
- The companies that calm markets first will likely be the ones that can show measurable workflow ownership (data rights, distribution, compliance posture), not just 'we use AI.'
And yes, it's a little surreal: the more AI 'works,' the more some categories get treated as optional.