Figure 2
The workflow is divided into Phase one, Phase two, Phase three, and Phase four. Phase one, Delphi projection development and formulation, begins with desk research including initial brainstorming, analysis of literature, and database research, alongside expert interviews with n equals 12 for verification and additional information. A first internal evaluation workshop conducts content review and functionality testing, producing 25 projections. Expert workshops with n equals 9 discuss and review the preliminary projections with participants from supply chain management. Evaluation of workshop results uses quantitative and qualitative assessment and reduces the set to 13 projections. An internal evaluation workshop performs the final review, followed by a pre-test with n equals 5, including content review, functionality test, usability test, and participants comprising two academics and three practitioners. Phase two, Delphi panel selection, includes a selection process with 3,579 invited experts, 111 survey starters, 70 completed surveys, and a final response rate of 1.9 per cent. Expert panel demographics for n equals 70 include 15.1 years of working experience in S C M, knowledge in S C M of 3.8, and knowledge in A I of 3.1. The S C O R-D S process includes Orchestrate n equals 23, Plan n equals 10, Source n equals 5, Order n equals 5, Transform n equals 6, Fulfil n equals 19, and Return n equals 2. Countries include Germany n equals 59, Great Britain n equals 6, European Union n equals 2, U S A n equals 2, and Other n equals 1. Phase three, Real-time Delphi survey, provides an opportunity to change estimations based on other participants' statements in real time. Projection one, Projection two, Projection three, and Projection 13 each include estimation by the E P-I-D approach and an option for adding comments and sentiment, with intermediate projections indicated by ellipsis. Phase four, Scenario development, begins with quantitative analysis of final Delphi results, stakeholder group analysis using S C O R-D S and country, and sentiment analysis based on S C M and A I experience. It continues with qualitative analysis of 1,216 written expert comments averaging about 17.4 comments per expert, syntax and content analysis, and analysis of expert comments for the dimensions E P, I, and D. The final scenario development applies fuzzy c-means clustering to classify projections into three clusters named Autonomous Operations, Integrated Intelligence, and Trusted Autonomy.Detailed process of the RTDS
Note(s): Five-point Likert scale; 1: Very low, 5: Very high, EP: expected probability of occurrence in percent; I: impact in case of occurrence (5-point Likert scale; 1: Very low, 5: Very high); D: desirability of occurrence (5-point Likert scale; 1: Very low, 5: Very high)
Source: Adapted from Steiner et al. (2024)