Artificial Intelligence and Machine Learning are making headways these days in many industries. PLM marketing materials are full of AI and ML buzzwords and promise how new tech will make it different. Analysts and researches are coming to outline PLM and AI opportunities. This is just one the many examples – When PLM Met Machine Learning: The Beginning of a Great Relationship (https://www.aberdeen.com/featured/blog-when-plm-machine-learning-meet/)

With the advent of big data, machine learning is introducing new ways to manage everything in manufacturing, from customer engagement, to design, to supply chain, to product lifecycle management.  The combination of big data and advanced, inexpensive computing systems has made machine learning practically in real work applications, and it is driving positive change as part of the product development lifecycle. Today’s Best-in-Class companies are embracing a new era of digital transformation in product lifecycle management (PLM). Among their quiver of Fourth Industrial Revolution tools is machine learning, which they are almost 2x more likely to implement than All Others (Figure 1).

While I like all these promises, I also think that something is missing. The usual pitch of ML sounds like this – PLM collect a lot of data and now we can use it for A, B, C. A typical things in the list are decision making, productivity, cost, etc… All these things must be improved anyway. To say using data is easy. Actually to use data, so it will bring results is actually much harder.

In such a context, I like to bring Amazon as an example of the company that started to use data a long time ago and developed impressive results. Recent Forbes article – The Twenty-Year History Of AI At Amazon can give you an idea of how Amazon is using data.

Read the article and draw your conclusion. My personal favorite is this one.

Machine learning is used throughout the entire customer’s search journey on the site through features such as predictive typing, optimising page layouts, or recommendations and suggested products. In fact, without machine-learning powered search, hyperpersonalized offers are not feasible. Hyperpersonalization is the AI-enabled concept of treating each person as an individual and not generally bucketing them into certain groups. Using machine learning, companies can develop a unique profile of each individual, and have that profile learn and adapt over time for a wide variety of purposes including displaying relevant content, recommending relevant products, website layout optimization and more. Amazon was one of the very early adopters of this pattern applying personalization and making recommendations across Amazon for the last twenty years. AI-enabled systems are increasingly blurring the lines of search and recommendation systems. By knowing which customer is searching for what product, AI will now make it possible to bring up very relevant results and personalized recommendations.

This example shows that AI and ML aren’t happening overnight. To learn how to use data, to find the right data and to figure out how to use it in specific applications and tools. In such a context, I found Siemens PLM NX experience with ML really interesting.

It made me think, to make AI and ML in PLM possible, the long program is needed. Such a program will need to build a model of product data, to bring product data out of existing legacy places, databases, storages; to connect data from silos and build use cases to empower future PLM features. Here are some of these possible features.

1- Generative part supply sourcing

We can see many examples of generative design these days, but 70% of components are actually not designed by the same company, but outsourced or purchased. To have a tool capable to plan sourcing strategy for global manufacturing can be an interesting opportunity. Think about Amazon for manufacturing…. well, it sounds a buzzword, unfortunately. But, the question of where to order parts and how much to pay is probably not.

2- Cost prediction

One big question for all time in manufacturing. What will be the cost? Cost prediction can get ugly and sometimes very not predictive. There are so many factors. Imagine a manufacturing company working for custom machine manufacturing. How to get 20 years of order history and predict the manufacturing cost of a new machine

3- XaaS – everything as a service

The manufacturing industry is moving to “everything as a service model” (XaaS). To be able to predict the cost of maintenance is becoming as critical as to know how much will cost to build a machine. So, from “nice to have, it is not my problem”, maintenance is becoming one of the most important IT requests. Try to figure out maintenance cost by analyzing the history of the data. Or try to get all the data together first.

What is my conclusion? PLM vendors and manufacturing companies should learn how to thread data. How to go beyond a “single version of truth” mantra. AI and ML in manufacturing is not a task for junior achievers bringing math club knowledge to manufacturing. It starts from the systematic use of data, building use cases and restructuring IT functions and PLM software to consume such data. Can be a tricky task and it can take time. 9 Women Can’t Make a Baby in a Month. Just my thoughts…

Best, Oleg

Disclaimer: I’m co-founder and CEO of OpenBOM developing cloud-based bill of materials and inventory management tool for manufacturing companies, hardware startups, and supply chain. My opinion can be unintentionally biased.

The post What PLM can learn from twenty years of Amazon AI appeared first on Beyond PLM (Product Lifecycle Management) Blog.