Serval's Jake Stauch on AI-Native Enterprise Automation and Control

Published 2026-05-19 - Runtime about 38 min - Watch on YouTube

Serval’s bet is that enterprise AI wins by making automation easier to create than the manual work it replaces. Jake Stauch argues the real moat is not raw model capability, but the control layer, customer intimacy, and a team small enough to keep reinventing the product as models change.

What Matters

• Serval is positioned as an AI-native ServiceNow: enterprise service management for employee support, where requests should resolve instantly instead of via tickets.
• The key design principle is inversion: building automation must be simpler than doing the manual task, or employees will just reset the password themselves.
• Serval uses two agents: an admin agent creates tools and permissions, while the help desk agent can only act through admin-approved workflows.
• Stauch says the product moat is boring enterprise controls: permissions, approvals, scoped API access, audits, reporting, logs, and alerts.
• Serval uses OpenAI models for end-user interaction and Anthropic models for code generation, then slows releases when new models break prompt-tuned behavior.
• The company is not optimizing cost first; it relies on reusable TypeScript-like automations, so common requests stop requiring fresh code generation every time.
• Stauch’s hiring mantra is fewer, better: he thinks talent density is the only durable moat because the product may need to be rebuilt every six months.
• He sees the central tension in AI adoption as autonomy versus control: employees want agents to do everything, while enterprises want guardrails.