NEWS
Bringing Speed and Certainty to Complex Product Configuration
Generative AI is useful in manufacturing, but companies need to marry it with determinism, according to a recent fireside chat with Laura Beckwith, Ph.D., Director of Product Management at Configit.
A product configuration for manufacturing is not valuable because it sounds right or looks plausible on a screen, Beckwith explains. It is valuable only if it can actually be engineered, sold, built, and supported.
In combination with determinism, generative AI fits best at the front of the process and across it, not just at the point of final validation.
Manufacturing does not run on “likely” or “probably.”
Generative AI works probabilistically. It predicts, infers, and suggests. In many business settings, that is acceptable. In manufacturing, it often is not. A configuration can look reasonable and still break engineering rules, compliance requirements, material constraints, or process limits. When that happens, the result is not a minor error. It becomes scrap, rework, delay, or a product that should not have been approved.
The core design choice is straightforward: use generative AI to interpret intent, and use deterministic logic to decide validity.
For example, instead of forcing a customer, seller, or engineer through a long sequence of option fields, GenAI can interpret plain-language requests and translate them into likely choices. That makes the experience faster and easier to use. In this case, GenAI is making complex products easier to configure, but determinism still needs to come into play before moving forward.
That logic layer matters more as product complexity increases.
A single feature choice can force a component change. That component change can affect weight, power, packaging, compliance, or downstream manufacturing steps. What looks like a small customer-facing decision may carry consequences across engineering, operations, and service.
This is also why configuration management cannot sit in one department. In many manufacturers, the rules are split across systems, spreadsheets, PDFs, and institutional memory. Engineering owns part of the logic. Sales applies another part.
Manufacturing works around constraints that never made it back into the formal model. While there is plenty of data, logic is fragmented.
Generative AI can help as complexity grows.
GenAI can read unstructured material, extract patterns, and propose rules that could be turned into formal configuration logic. That can reduce manual effort and surface dependencies buried in legacy documents.
Even so, any AI-generated recommendation still needs to pass through a deterministic model that shows whether the result is actually allowed and what the downstream impact will be. That is what lets teams move faster without giving up control. It also keeps accountability where it belongs.
GenAI can reduce translation work. It should not quietly become the authority.
The human role does not disappear in this model.
AI can generate options quickly. Deterministic logic can verify what is valid. People still have to judge what makes sense for the business, the customer, and the operation.
That matters because AI systems are often persuasive even when they are wrong. So “human in the loop” is not enough on its own. Manufacturers need workflows that make validation visible: what the system suggested, why it suggested it, and which rules it triggered.
The practical starting point is product logic, supported by GenAI. Find where configuration knowledge lives, where it is fragmented, and where teams are recreating the same rules across systems. Then decide where GenAI can reduce friction or speed up model creation without becoming the source of truth.
In manufacturing, that is the real boundary. AI can guide the process. Deterministic logic has to govern the result.
Ready to learn more? Watch the full webinar, GenAI Meets Determinism: Bringing Speed and Certainty to Complex Product Configuration.
NEWS
