Motivation – Automation can fail to deliver the target safety or productivity benefit as intended by those managers and designers advocating its introduction. In a safety critical domain this problem is of significance not only because the unexpected effects of automation might prevent its widespread usage but also because they might turn out to be a contributor to incident and accidents. Originality/value – Research on failures of automation to deliver the intended benefit has focused mainly on human automation interaction. This PhD research plan aims at characterizing decisions - taken under productive pressure - for those involved in the automation development process, to identify where and when the initial intention the automation is supposed to deliver can drift from the initial idea. Expected Finding – The objective is to develop Anti-Drift Principles to identify and compensate proactively for possible sources of drift in the development of new automation. Research Approach – The research is based on case study and is currently entering Year 2.