This article is part of the How Do YOU CM2? blog series in collaboration with the Institute for Process Excellence (IpX). Although I receive compensation for writing this series, I stand behind its content. I will continue to create and publish high-quality articles that I can fully endorse. Enjoy this new series, and please share your thoughts!
Two 2024 studies map where AI is landing in engineering change management. The pattern is worth examining.
A systematic literature review covering 39 AI applications in ECM (Burggräf et al., Research in Engineering Design, https://link.springer.com/article/10.1007/s00163-023-00430-6) found the field converging on two capabilities: predicting change propagation (what else will this change affect?) and knowledge retrieval (what do we know about similar changes?). A separate 2024 study in the International Journal of Production Research (https://www.tandfonline.com/doi/full/10.1080/00207543.2024.2432465) addressed a further downstream problem: using machine learning on automotive OEM change data to predict adherence to effectivity dates, whether a planned cut-in commitment will actually hold up against production reality.
Both are genuine engineering problems. Change propagation prediction reduces missed impacts. Effectivity date adherence prediction surfaces execution risk earlier.
What they share: both operate downstream of the scope confirmation step.
In a properly sequenced change process, applicability, which products and variants the change applies to, is confirmed before effectivity is set. That confirmation is the precondition for the impact assessment and the baseline against which any cut-in commitment is made. The research shows that AI is developing significant capabilities at the execution and impact layers. That’s where the data is richest and the tooling most mature.
The upstream step, whether scope was formally confirmed before effectivity was committed, is harder to encode in a model. It requires a documented process event that many organizations never produce. You cannot train on data that doesn’t exist.
CM2’s Enterprise Change Assessment phase is designed to produce exactly that: a documented applicability baseline, which configurations are affected, and to which of them the change applies, before any CIB planning begins. For organizations building toward AI-assisted ECM, that documented upstream confirmation is the process signal that makes downstream modeling meaningful. Without it, the model is operating on inputs whose validity was assumed rather than confirmed.
Practitioners deploying AI in ECM should ask: what does this tool assume about the quality of the process inputs it’s operating on?
What AI applications is your organization exploring in change management, and at what stage in the process are they targeted?