The AI Workforce Buildout
Camilo Serna Zamora@tekelala
−1.5%
since Sep 30
The idea
I believe companies will soon run AI agents alongside human employees, hiring, firing, and managing them like a new kind of workforce, and adoption inflects over the next couple of years. This basket owns the platforms building and shipping those agents, plus the identity, security, and monitoring tools needed to keep them in check, since every AI worker needs its own credentials and guardrails. It leans concentrated and higher-beta rather than broad, with a small cash reserve set aside to add on dips. If adoption stalls or agent products stop shipping, the thesis is wrong.
What it owns
- These platforms are creating and selling AI employees that companies can hire to do software work. · MicrosoftBuilds Copilot agents that get assigned work inside tools companies already use, collecting the wage bill as adoption grows., ServiceNowStaffs its operational workflows with AI agents that complete tickets instead of routing them to humans for customers to pay for., SalesforceSells Agentforce to companies as AI employees by the conversation for sales and service jobs, with revenue tied to agent deal volume., PalantirPlaces AI agents directly on company operational data so they act on real decisions, not just chat, accelerating commercial customer growth., AlphabetBuilds Gemini models and agent tooling that companies use to create AI workers, getting paid whether they build their own or rent Google's., SAPPuts AI agents into finance, HR and supply chain work across its back-office customers, growing cloud revenue as they activate agents.38%
- Every AI worker needs credentials and access rights, and these companies manage that system for non-human employees. · OktaProvides login systems for non-human workers, opening a new product line as companies buy identity seats for agents alongside people., ZscalerExtends its zero-trust model to AI agents making thousands of requests daily, growing as it lands agent-security products., SailPointGoverns what each worker is allowed to access and handles revoking permissions when AI employees are fired from jobs.17%
- AI employees that get tricked or misbehave create security risks, so platforms that police software behavior become more critical. · Palo Alto NetworksShips security products specifically designed to contain AI agents before they go rogue, creating a new budget category for companies., CrowdStrikeIt monitors what processes and applications actually do at runtime, catching unauthorized or anomalous behavior that could signal an AI agent acting outside its intended scope., CloudflareActs as the tollbooth for AI agents browsing, calling APIs and working, capturing traffic volume as agent proliferation grows.16%
- Someone must verify that AI employees are doing their jobs correctly, and these tools monitor software at work. · DatadogMonitors whether AI employees are doing their jobs correctly, with AI observability products becoming the performance review for software workers., DynatraceWatches complex software systems end to end, becoming more valuable as AI agents make those systems messier and riskier.11%
- AI workers run on chips and company data, so suppliers of both get paid regardless of which agent platform succeeds. · NvidiaSupplies the chips that run AI agents, with inference demand for operating agents rather than just training models sustaining growth., SnowflakeStores the company data that makes AI workers useful and is building tools for agents to query it directly, growing consumption from agent workloads.9%
- Cash10%