AI change management: A human-centric approach

The Cohere Blog article explains that traditional change management, which has focused on new tools or leadership shifts, is becoming inadequate as AI is integrated into everyday work. AI is not merely software to be deployed; it is an active participant that must be onboarded with clear roles, responsibilities, and workflows. The article highlights that AI’s open‑ended, context‑dependent nature requires human‑in‑the‑loop mechanisms, allowing AI to request missing information or defer to human judgment when necessary. Employees transition to AI managers, setting direction, assessing quality, and deciding when human input is needed, while avoiding the risk of “AI slop” by prioritizing impact over volume. For those building internal AI solutions, the role expands to include product‑management responsibilities such as defining user needs, business outcomes, and governance across the lifecycle. Effective collaboration depends on distinguishing tasks that can be delegated to AI from those that require human judgment, a division that must be continually reassessed. Governance for AI must be more adaptive than for traditional software, with ongoing monitoring for performance drift, bias, and alignment with organizational values. Access controls should grant AI least‑privilege permissions, allowing it to request additional access under approval. The article draws parallels between AI governance and people management, suggesting that AI should be able to explain its outputs and be subject to regular check‑ins and course corrections. Governance should be tailored to the autonomy level of each AI system, from human‑operated tools to fully autonomous, always‑on systems, ensuring appropriate real‑time monitoring and human oversight for high‑risk decisions.