IT Admin for the AI Workforce — Sarthak Aggarwal, Decawork
AI Engineer World's Fair 2026 · 16:17
← All organizationsdecawork.ai ↗
AI agent governance and infrastructure
Decawork
Decawork builds an agent control plane for IT teams to deploy, govern and maintain internal AI agents built across different tools. Teams can keep their chosen development tools while IT manages agent ownership, access and approvals in one place. The platform records attributable actions, monitors runs, alerts IT to failures or suspicious activity, and retires unused agents. It addresses the operational work of turning employee-built agents into company-managed systems.
Founded in 2026, Decawork is led by cofounder and CEO Aman Raj and cofounder and CTO Sarthak Aggarwal. Raj previously built an AI compliance platform at Barclays; Aggarwal built AI systems at NVIDIA. Decawork’s technical approach gives each agent its own identity and task-scoped access. Agents do not hold raw credentials: tool calls pass through a gateway that logs actions and requires approval for sensitive operations.
The San Francisco company joined Y Combinator’s Summer 2026 batch. Its accelerator profile listed a team of two as of August 2026.
1 talk
Newest first1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Messages from the stage
Separate planning from untrusted content
Aggarwal proposes separating trusted planning from untrusted content processing, with explicit capabilities and enforceable policy constraining execution. His security examples illustrate why instructions and model guardrails are insufficient boundaries.
Affiliations reflect each recorded session, not necessarily current employment.