Agentic AI governance must be built in from the start, not added later

Roman Leliukh and Patrick Pichler argue that AI governance, like safety measures for past technologies, cannot be retrofitted after deployment. They cite historical examples where regulations for steam engines, cars, and airplanes only followed widespread use. The authors emphasize that AI agents, which act like coworkers with system access, require integrated governance from day one. This governance must provide clear identity, controlled authority, and audit trails for every action. Their company, Kimchi, aims to establish this necessary framework proactively.