Manufacturing ERP vendors are locked in a branding contest. SAP, Epicor, QAD, Syspro, IFS, Oracle: each has announced some version of agentic AI over the past twelve months. The press releases promise autonomous decision-making, self-correcting supply chains, predictive everything. The technology teams we place into these environments describe something more modest. Most implementations remain at the copilot stage, surfacing recommendations that a human still approves.
We track this gap closely because it shapes who our clients hire. A manufacturer in Bavaria recently rewrote a senior SAP role mid-search. The original specification called for experience deploying AI-driven planning tools. After three weeks of internal review, the requirement softened to experience evaluating AI-driven planning tools. The first version assumed production-ready functionality. The second assumed a pilot that needs adult supervision.
Companies want candidates who can separate vendor marketing from deliverable capability. The skill is part technical, part commercial: reading a product roadmap, stress-testing a demo, asking the questions that reveal whether the automation actually runs unattended or whether it needs a human backstop. That evaluation skill used to sit with consultants brought in for a project. Now it appears in permanent role specifications for SAP programme managers and operations leads.
If you have run a genuine agentic deployment, even a narrow one, document precisely what it did without human intervention and what it could not. That specificity cuts through the noise in every interview. If you are hiring for these roles, ask candidates to walk you through a vendor claim they rejected and why. The answer reveals whether they can protect you from a roadmap that arrives two years late.
Prompted by reporting from ERP Today.