Heather Sye’s AI maturity model exposes the gap between experiments and operating-model change
From AI tools to governed decision architecture
Heather Sye published a LinkedIn post framing AI maturity in five stages: casual experimentation, AI as a task tool, AI as workflow automation, AI as infrastructure, and the self-healing organization. The post is useful because it separates visible AI activity from real operating-model change: using ChatGPT or Copilot is not the same as redesigning processes, data flows and decision ownership.
The strongest Supply Chain signal is the gap between Stage 2 and Stage 3. Many organizations can point to isolated productivity gains: summarizing RFPs, flagging supplier delays, drafting reports or automating analysis. Fewer have reached Workflow Automation, where repetitive handoffs are removed, exceptions are routed to the right people, and the process itself starts to change.
Stage 4, AI as infrastructure, is where the maturity model becomes more demanding. Connected systems, shared data, immediate disruption visibility and automated decision support require more than pilots. They require AI Governance, data ownership, system integration, exception rules and clear accountability for what happens when a recommendation affects service, inventory, capacity or procurement exposure.
The final stage, described as the self-healing organization, is deliberately provocative. It imagines an organization where AI monitors AI, playbooks regenerate, and systems diagnose and correct failures. For Supply Chain leaders, this raises an important caution: self-healing operations may sound attractive, but they also require visibility, auditability, human override rights and boundaries around surveillance, control and decision accountability.
The practical value of Sye’s post is that it gives leaders a simple maturity mirror. The question is not whether teams are “using AI,” but whether AI has changed the decision system: what data is connected, which workflows have been redesigned, who owns the logic, and how risks are escalated when automated recommendations are wrong.
Dataleo angle
This insight fits the Radar because Supply Chain AI maturity is often overstated. The difference between a useful AI tool and an AI-enabled operating model is governance: data quality, ownership, validation, version control, access rights, manual override and traceability.
For planning, procurement and operations leaders, the maturity journey should be assessed by decision impact. Which decisions are faster or better? Which handoffs disappear? What is connected to APS, ERP, TMS or BI? Where does human judgment remain mandatory? A self-healing organization should not mean unmanaged automation; it should mean a governed Decision Architecture with clear failure modes and accountability.