Where AI adds value in Engineering Change Management — and where it does not
AI in Engineering Change Management will undoubtedly become increasingly important. But that does not mean that AI should be used everywhere.
The real opportunity is not to let AI take over Engineering Change Management. It is to use AI to remove low-value work, make complex information easier to understand and give engineers and other specialists more time for what they are actually needed for: understanding problems, developing good solutions and making informed decisions.
This requires us to distinguish between three things:
- Facts that already exist in our enterprise systems
- Information that requires interpretation and context
- Decisions for which people must remain responsible
AI can be extremely valuable for the second. It should not replace the first or the third.
Give engineers more time for engineering
Many Engineering Change Requests start with surprisingly little information.
“Bearing fails.”
“Customer requests another material.”
“Drawing needs to be updated.”
The person reporting the problem may understand perfectly well what they mean, but others who subsequently have to evaluate the change may not.
This is an obvious place for AI.
AI can help turn an initial observation into a clearer Change Request. It can improve the description, identify ambiguities and point out information that appears to be missing.
What is the actual problem? Where was it observed? Which product or installation is affected? What is the expected outcome?
The purpose is not for AI to decide what should be changed. It is to reduce the time people spend rewriting, structuring and clarifying information — and improve the information available to the people who will make that decision.
That is a much more attractive use of AI in Engineering Change Management: less administration, more engineering.
Make complexity easier to understand
Engineering changes accumulate information.
Comments are added. Alternatives are discussed. Decisions are made. Documents are attached. Questions are answered. New information appears as the change progresses.
Understanding a complex ECO may eventually require reading through pages of information and a long history of discussions.
AI can make this much more manageable.
It can summarise what is being changed, why the change was initiated, which major decisions have already been made, what remains unresolved and where concerns have been raised.
It can also help retrieve experience from previous changes. ECM systems contain valuable organisational knowledge about earlier problems, solutions, decisions and lessons learned. AI can make that experience much easier to find and reuse.
Again, however, AI should provide input to the decision, not make the decision itself.
An AI may identify that a similar change previously caused production problems. It may highlight an inconsistency or suggest a question that has not been considered.
But deciding whether to proceed, which solution to choose, when to release it and what risks are acceptable remains a human responsibility.
A BOM is not a language problem
There is another boundary that is just as important.
Imagine that an engineer is considering changing material 4711 and asks:
Which products use this material?
You could provide thousands of BOM records to an AI model and ask it to work out the answer.
But why would you?
If the relationships are maintained in PDM, PLM or ERP, the answer already exists. A database query or where-used analysis can determine it explicitly.
The same applies to questions such as:
- How many pieces do we currently have in stock?
- Which purchase orders contain the affected material?
- What quantities are expected to arrive?
- Which customer orders are affected?
- Which released BOMs contain this component?
These are not AI problems.
They are queries against structured enterprise data.
If the ERP system says that there are 1,247 pieces in stock, we need the answer 1,247. We do not need AI to infer what the answer is likely to be.
Facts should remain facts
This becomes particularly important when engineering changes are evaluated.
An organisation may need to decide whether to implement a change immediately or wait until existing inventory has been consumed. Stock, purchase orders, demand and product structures may all influence that decision.
The underlying facts should come directly from the systems that own them.
Using AI to reconstruct those facts introduces uncertainty where none is necessary. The result may be harder to reproduce and audit, and it may become unclear exactly where a number or relationship came from.
AI is extremely good at dealing with ambiguity.
There is little reason to introduce ambiguity where none exists.
AI is not free — financially or environmentally
There is also a more practical reason not to use AI for problems that conventional technology already solves well: AI consumes computational resources.
Analysing thousands of BOM records, stock movements or purchase-order lines with an AI model may require substantial processing and large numbers of tokens. A database can often answer the same factual question directly, faster and at a fraction of the computational cost.
Token prices may seem low today, but companies should be careful about designing processes that depend on continuously sending large volumes of enterprise data through AI models. As AI use scales across the organisation, unnecessary processing can become a real operating cost.
The same applies to sustainability.
Companies increasingly measure and reduce the energy consumption and carbon footprint of their digital infrastructure. AI should be part of that discussion. If a deterministic query can provide an exact answer with far less computation, using AI instead is difficult to justify.
This does not argue against AI. It argues for using it where its additional intelligence creates additional value.
Use the simplest reliable technology for the problem — and save AI for the problems that actually require intelligence.
Let AI explain the facts — not invent them
This does not mean AI has no role when working with structured enterprise information.
Quite the opposite.
The stronger approach is to combine reliable data with AI.
Let the enterprise systems establish the facts:
Material 4711 is used in 17 products.
4,820 pieces are currently in stock.
Two purchase orders are open.
Three affected products have current customer orders.
AI can then combine these facts with the change description, previous experience and other relevant information and help the engineer understand the situation.
It might highlight that existing inventory appears to be an important consideration. It might suggest examining interchangeability or comparing alternative phase-in dates.
But there is an important word here:
Suggest.
AI can prepare, organise, summarise, explain and challenge.
People decide.
The engineer decides whether the proposed technical solution is appropriate. The relevant business functions decide how the change should be implemented. And authorised people approve and release the change.
These decisions carry responsibility and consequences. They should not disappear into an algorithm.
Better engineering, not automated engineering judgement
The greatest potential of AI in Engineering Change Management may therefore be less dramatic than the vision of autonomous AI agents making engineering decisions.
But it may also be much more valuable.
Engineers spend significant time searching for information, reading discussions, preparing descriptions, summarising changes and communicating decisions. Every hour saved on these activities is an hour that can potentially be spent understanding the problem, evaluating alternatives and developing a better solution.
That should be the objective.
Not fewer engineers making fewer decisions.
Better-informed engineers with more time to make good decisions.
The future of AI in Engineering Change Management should therefore combine three strengths:
- Reliable enterprise systems provide the facts
- AI helps people find, structure, interpret and understand information
- People apply engineering knowledge, business judgement and accountability to make the decisions
That is a much stronger combination than asking AI to do all three.
Use AI to support human intelligence — not replace it.
And don’t use AI to guess what your enterprise systems already know.
Let’s talk about AI in Engineering Change Management
Where can AI genuinely improve your ECM process — and where should reliable enterprise data and human judgement remain in control?
BoostPLM can help you apply AI where it adds real value, supports better decisions, reduces administrative work and frees more time for engineering.
Talk to us about AI in ECM