Agentic AI: Who Eats Whose Lunch — A Concrete Winner/Loser Map

Winners: Infrastructure for Agents

Agents need infrastructure to operate: compute for inference, memory for context, data platforms for knowledge retrieval, and networking for coordination between agents. The companies providing this infrastructure are in Natural Position — they benefit regardless of which agents or which use cases win.

• Cloud compute: AWS, Azure, GCP — agents run on their infrastructure
• Data platforms: Snowflake, Databricks — agents need data to act on
• Observability: Datadog, Elastic — monitoring agent behavior at scale
• Semiconductor: The entire 6-layer supply chain benefits from more inference demand

Losers: Workflow Replacement

Companies whose core product is automating a human workflow face existential competition from agents that automate the same workflow without the software layer. The SaaSpocalypse maps this in detail. Key vulnerable categories:

• Tier-1 customer support: Zendesk, Intercom — agents handle routine tickets better and cheaper
• Basic CRM: Low-end Salesforce use cases — agents manage contacts and follow-ups natively
• Project management: Monday.com, Asana — agents coordinate tasks without a dashboard
• Expense management: Concur, Expensify — agents process receipts and approvals automatically

The Uncertain Middle

Cannibalize or Die companies have the resources to pivot but must destroy their own revenue model to do it. Microsoft is the clearest example: Copilot cannibalizes Office per-seat revenue but positions Microsoft as the agent platform. Salesforce with Agentforce faces the same dynamic. The investment thesis on these companies is pure execution: are they moving fast enough?

Key Companies