Company does not have processes. It has habits (ai.levelbrook.com)

In a landscape increasingly populated by companies automating business processes with AI, the emphasis has shifted from capability to understanding the underlying processes within organizations. The article highlights that many businesses operate without clear, documented processes; instead, they rely on unrecorded oral traditions and informal practices that can lead to inconsistencies and errors. This lack of formal documentation reveals a hidden opportunity for AI automation projects: creating accurate specifications of processes can enhance onboarding, streamline operations, and expose inefficiencies or contradictions that previously hindered performance. As AI tools become more integral in managing operations, the article argues for a paradigm shift in how these systems are designed. Rather than aiming for full autonomy where human oversight is eliminated, the focus should be on making human decision-making more efficient and less costly. Key architectural choices include ensuring models make proposals without holding credentials, maintaining a closed vocabulary for actions, and implementing robust accountability through clear documentation and a structured decision-making ladder. Ultimately, the ledger of decisions and actions becomes the product, enabling better auditing and fostering trust in automated systems.