Manufacturers considering investment in AI-driven Advanced Planning and Scheduling (APS) must commit capital before any internal track record exists to appraise it against, a condition classical methods such as Return On Investment (ROI), Total Cost of Ownership (TCO), and Time-Driven Activity-Based Costing (TDABC) were not designed to accommodate. This thesis develops and evaluates the AI-APS Investment Evaluation Framework, extending TDABC to price a known production plan against known demand rather than a completed period, structured around four AI-specific constructs, a causal chain linking AI-APS capability to priced waste, and deterministic operating levers in place of stochastic simulation. The framework is applied to a pilot manufacturing case, comparing a Client's historical actuals under a rule-based scheduler against a Vendor's AI-optimised scheduler's priced plan for the same real, known demand. The comparison shows the scheduler's plan generating roughly 31% less waste per euro of value-adding work, a ratio of 0.83 against a historical 1.20, and a corrected improvement of €2,411.20 per working day once the cost of the additional output itself is properly accounted for. Because historical idle time and scheduler waiting time are not equivalent concepts, and because the scheduler's plan produces substantially more output per day than the historical baseline, this improvement rests almost entirely on the Cost of Waitings and should be read as directional rather than precise. An illustrative appraisal, using assumed subscription, implementation and datareadiness costs, shows that the case is governed by the share of the projected saving that materialises, not by cost: NPV stays positive as long as roughly 4% to 7% of the projected saving is realised. Lever-scenario analysis further shows this saving is more sensitive to operator availability than to machine breakdowns or lateness, echoing the wider literature's finding that AI value realisation is conditional on organisational capability, not investment size. The result is a priced, verifiable comparison of a known plan against a known baseline, with an illustrative, assumption-driven appraisal, not a validated prediction of operational performance or a confirmed business caseThis thesis develops and evaluates an ROI calculation framework and accompanying calculator for AI-driven APS investments in manufacturing. Following the Design Science Research Methodology (Peffers et al., 2007), the study first synthesizes established ROI/TCO conventions, emerging AI-specific ROI literature, and Overall Equipment Effectiveness (OEE) research to identify which framework elements transfer to the AI-APS context and which require new constructs. On this basis, a conceptual framework is built that traces a causal chain from AI-APS capabilities through OEE component improvements to operational effects and financial value drivers, and defines the minimum viable set of data inputs exportable from existing APS systems. The framework addresses the absence of a pre-implementation baseline through scenario-based estimation (conservative, base, optimistic) combined with one-way sensitivity analysis, producing ROI ranges rather than point estimates. The framework is applied to real client data provided by Quindi, where predicted returns are benchmarked against observed post-implementation performance. The thesis contributes theoretically by adapting classical ROI constructs to AI-specific investment characteristics, and practically by delivering a documented, reusable calculator for practitioners. Limitations include the single-case design and the deliberate use of deterministic scenario analysis rather than stochastic simulation, both identified as avenues for future research.
Investing Without a Baseline: An Appraisal Framework for AI-Driven Advanced Planning and Scheduling in Manufacturing
BARANOWSKI, ARTUR SLAWOMIR
2025/2026
Abstract
Manufacturers considering investment in AI-driven Advanced Planning and Scheduling (APS) must commit capital before any internal track record exists to appraise it against, a condition classical methods such as Return On Investment (ROI), Total Cost of Ownership (TCO), and Time-Driven Activity-Based Costing (TDABC) were not designed to accommodate. This thesis develops and evaluates the AI-APS Investment Evaluation Framework, extending TDABC to price a known production plan against known demand rather than a completed period, structured around four AI-specific constructs, a causal chain linking AI-APS capability to priced waste, and deterministic operating levers in place of stochastic simulation. The framework is applied to a pilot manufacturing case, comparing a Client's historical actuals under a rule-based scheduler against a Vendor's AI-optimised scheduler's priced plan for the same real, known demand. The comparison shows the scheduler's plan generating roughly 31% less waste per euro of value-adding work, a ratio of 0.83 against a historical 1.20, and a corrected improvement of €2,411.20 per working day once the cost of the additional output itself is properly accounted for. Because historical idle time and scheduler waiting time are not equivalent concepts, and because the scheduler's plan produces substantially more output per day than the historical baseline, this improvement rests almost entirely on the Cost of Waitings and should be read as directional rather than precise. An illustrative appraisal, using assumed subscription, implementation and datareadiness costs, shows that the case is governed by the share of the projected saving that materialises, not by cost: NPV stays positive as long as roughly 4% to 7% of the projected saving is realised. Lever-scenario analysis further shows this saving is more sensitive to operator availability than to machine breakdowns or lateness, echoing the wider literature's finding that AI value realisation is conditional on organisational capability, not investment size. The result is a priced, verifiable comparison of a known plan against a known baseline, with an illustrative, assumption-driven appraisal, not a validated prediction of operational performance or a confirmed business caseThis thesis develops and evaluates an ROI calculation framework and accompanying calculator for AI-driven APS investments in manufacturing. Following the Design Science Research Methodology (Peffers et al., 2007), the study first synthesizes established ROI/TCO conventions, emerging AI-specific ROI literature, and Overall Equipment Effectiveness (OEE) research to identify which framework elements transfer to the AI-APS context and which require new constructs. On this basis, a conceptual framework is built that traces a causal chain from AI-APS capabilities through OEE component improvements to operational effects and financial value drivers, and defines the minimum viable set of data inputs exportable from existing APS systems. The framework addresses the absence of a pre-implementation baseline through scenario-based estimation (conservative, base, optimistic) combined with one-way sensitivity analysis, producing ROI ranges rather than point estimates. The framework is applied to real client data provided by Quindi, where predicted returns are benchmarked against observed post-implementation performance. The thesis contributes theoretically by adapting classical ROI constructs to AI-specific investment characteristics, and practically by delivering a documented, reusable calculator for practitioners. Limitations include the single-case design and the deliberate use of deterministic scenario analysis rather than stochastic simulation, both identified as avenues for future research.| File | Dimensione | Formato | |
|---|---|---|---|
|
Baranowski_Artur_Slawomir.pdf
accesso aperto
Dimensione
2.36 MB
Formato
Adobe PDF
|
2.36 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/118986