What Manufacturing Leaders Misjudged About AI in 2025
In 2025, many manufacturers implemented artificial intelligence across forecasting, logistics, and supplier risk scoring, expecting breakthroughs. By year-end, though, a pattern had emerged: AI improved awareness and decision support, but it did not eliminate uncertainty or deliver automatic resilience.
- AI doesn’t replace judgment — it augments it
Procter & Gamble’s Jamie McIntyre Horstman explained that machine learning models significantly enhance demand forecasting by identifying patterns like seasonality and removing outliers. Models such as “long-term demand sensing” and daily forecasting suites help reduce error. But these outputs are not definitive predictions; they are probability-informed trend estimates that require human interpretation. Planners still decide how to act on them — especially when models signal conflict or uncertainty.
AI augments decision-making but does not remove judgment from the process.
- Missing and conflicting data still require human intervention
AI models, by design, depend on data quality. When data is incomplete or conflicting — such as in emerging geopolitical signals or first-time market behaviors — AI outputs can be misleading without context. Horstman described how teams use regression analysis and keyword patterns to interpret unusual demand behaviors like panic buying — situations where historical patterns don’t yet exist.
The session reinforced that AI is as strong as the data that feeds it, and when that data lacks breadth or clarity, humans must fill the contextual gaps.
- Supplier risk scoring cannot fully automate risk avoidance
Srinivasan Narayanan recounted how traditional supplier risk assessments were quarterly and reactive. AI now continuously monitors delivery performance, financial signals, and external indicators. However, AI does not automatically solve supply risk. It surfaces early warnings — and manufacturers still decide how to respond through actions such as dual sourcing, inventory adjustments, or negotiations.
The expectation that AI alone would prevent disruptions proved overly optimistic; it is effective mainly as an early warning system integrated into decision workflows.
- Internal data sharing remains a constraint
Maria Araujo highlighted that visibility into deeper supply tiers remains limited. While AI can correlate internal signals with external indicators, lack of shared data outside direct suppliers remains a barrier. Manufacturers often hesitate to share sensitive operational information — even if doing so could improve predictive insights across multi-tier networks.
Misjudging the role of data governance and cooperation has slowed the realization of AI’s full potential.
Concluding reality in 2025
Across industries in 2025, AI matured from pilot projects to embedded operational tools. Yet, as the panel made clear, the promise of AI has not materialized into fully autonomous supply chain operations. Instead, manufacturers now recognize that:
- AI provides context and early signals, not answers.
- Human judgment remains central to decisions.
- Data sharing constraints limit deeper predictive power.
- AI systems improve situational awareness — not full certainty.
In sum, the industry’s misjudgment was not overestimating AI’s utility — it was assuming it could replace the very human decisions that still drive supply chain resilience.
This article is based on the session “Can AI Predict and Prevent the Next Supply Chain Disruption?” at IIoT World Manufacturing & Supply Chain Day 2025, organized by IIoT World.