This content relates to : DIGITAL TRANSFORMATION
Robert G. Cooper
McMaster University
Hi ladies and gentlemen, this is Bob Cooper and I’m going to talk a little bit about the role of AI in new product development. In other words, how AI can help product developers come up with new products better, faster, cheaper, more effectively and so on. Artificial intelligence has been defined technically and I’m sure you’re familiar with some of the definitions, but I heard a definition that was very well said and made it much easier for me to understand.
The definition is that AI is a prediction tool that is simply more efficient and more cheap or cheaper than any other prediction tool you can imagine. It’s a prediction tool, a very effective, very cost effective prediction tool. I had to think about that.
Well, I could understand doing market forecasts using AI. Of course, AI can recognize patterns that normal statistical analysis could not and do a much better market forecast, I would imagine. But he went on to say, where your problem is not one of forecasting, what AI does is it translates that problem into a forecasting problem.
Or it may say, let’s say you’re a designer of hot wheel cars at Mattel. The designer in AI basically says, how would a good designer design a car to meet the following requirements? It predicts how the designer or how would Picasso paint this mountain range scene in the south of France? It predicts how or how would a good driver drive this Tesla, this autopilot Tesla to the next intersection, given that there are six people walking across the road, what would a good driver do at this point? It’s predicting all the time. So that’s a very interesting way.
This came out of a book that was written by Agrawal out of Harvard. I can’t remember the name of the book, but you might want to check him out. Good book. It’s on Amazon.
You’ll see it. Well, one of the things when I started getting into this area of what can AI do for me as a product developer or student of product development, one of the things I found is there are more than 40 different applications I found online. And this is largely through a search of vendors.
Vendors are very, very visible online. They got lots of blogs and lots of websites. And I counted 40 unique applications from, I think there was more than 500 vendors I came across.
And not just the IBMs and the Microsofts and the Googles, but lots of smaller companies as well, specialty suppliers. And so I had great difficulty sort of coping was overwhelming this landscape. So I thought landscape, I’m going to come up with a map.
And so we came up with a map and across the top of the map or the east-west coordinates, we put the stages of the stage gate process. Idea through to commercialization or launch, and finally post-launch. And down the left-hand side, oh, that’s location in the new product process.
Down the left-hand side, we put the role of AI and different authors have talked about the different roles that AI has. Sometimes AI is just a facilitator. For example, it helps you, the user do a better market forecast.
It basically does what you do only a little bit faster, a little bit cheaper. On the other hand, AI up at the top of the screen can also be an originator, generate ideas, for example, that you could not do. It can become very creative, far beyond what most humans can do at this point,
and that capability is increasing, increasing, increasing, as you know. So based on this X, Y or north, south, east, west axes, we created a map. Here’s the front end of the new product process.
Here’s the back end of the new product process. And here are some, but not all, of the 40 applications we found for AI and product development. And I must admit, I’ve investigated most of these and found out they are legitimate.
And I’ve written a number of articles about the different types of applications. So this is somewhat overwhelming for an AI task force who wants to try to install and modernize their new product development process. It’s a little overwhelming, but at least this map helps.
Some of the abbreviations down in the green box on the right. So where does it apply? Well, let’s talk about the front end, the front end of the process, ideation, concept development, concept refinement, concept testing. Here are about some of the tools.
ChatGPT can generate very novel ideas. In fact, there’s been several research studies done to show that the ideas that ChatGPT comes up with are better than what humans come up with. They put them in match the sort of competition here with a judging panel.
Try it. Put in some prompts like, I would like to come up with some ideas for a new Hot Wheels car aimed at my granddaughter or my new recently born daughter. She’s eight years old and wants to play with Hot Wheels, but there’s no feminine type Hot Wheels.
What would you come up with? Ask it. It’ll tell you. Incredible.
I’ve done this in companies and they’re stunned by the ideas that ChatGPT comes up with. And of course, there’s more sophisticated versions of idea generation software that you have to pay for. The ChatGPT I’m using is free.
And the more sophisticated ones are even better, obviously. Another thing that AI will do is scan the internet looking for unstructured text to identify market gaps and emerging customer needs. Example was done by Advanced Market Science, a market research firm in the Boston area, Boston, USA, on basically blood tests that measure the glucose levels of people with diabetes.
They went online and found thousands, tens of thousands of comments on blogs about mostly written by unhappy people with their testing equipment and took all this information, scanned it, made sense out of it, broke it down into fundamental needs, which were then tested. And then from that, the researchers were able to come up with a set of requirements for the ideal new product, the one that would address the pain points of the typical customer. But no human could have done that.
Taking tens of thousands of comments and reducing them down to a handful of themes. These scanning blogs are commercially available. They’re online.
A lot of them are user comments, scanning user comments, complaints. I was doing some work for a company that made tractors and farm equipment and basically posed a question. How many online comments do you think farmers put online about tractors or their farm equipment? They had no idea.
They’d never checked it. I checked it later on. The number I got was something like three million.
Three million farmers had gone online to make comments about their tractors and what they wanted in them. What an opportunity to do voice of customer research. Another thing that AI can do is analyze responses from surveys and interviews.
Lots of people do interviews. They try to transcribe them, personal interviews. They try to transcribe them.
It’s in text. It’s unstructured text. Hard to handle that.
AI does a marvelous job of that, natural language processing, NLP. It’ll generate a first draft of customer interview guide. It’ll generate new product concepts from these scanning insights.
It’ll generate them into various concept proposals. Nestle is doing that with beverages. In fact, Nestle has taken that approach even further and doing concept testing online using AI.
Not sure how they do that. That’s a little proprietary, but that is possible. There’s also public companies or companies that are operating publicly that are offering that service as well.
Here’s another thing Nestle is doing. They’re mining internal technical data. They got a lot of research in the food industry, food and beverage, to find out opportunities for exploitation.
They’re reaping tens of millions of dollars from research that sits in a report sitting on some shelf, figuratively speaking, sitting in some computer file, unused. Another thing AI will do is design and draw the product concept from verbal commands and predict customer reaction to them. General Motors is doing this now.
In the old days, artists would draw the new concept car, perhaps using CAD, computer-aided software, computer-aided drawing, and then they would invite a whole bunch of customers in to see the drawings or the 3D sketches translated into 3D drawings, and they’d get their reaction. Now, working with MIT, they’re actually using AI to draw the car based on verbal commands, make it a little this or make it a little longer, make it look a little more aggressive, make it a convertible, et cetera. The software is able to predict customer reaction to that concept car.
Do they like it? Do they think it’s novel? This is saving months of time and hundreds of thousands of dollars per test. Another thing that AI is able to do is do multiple iterations of the concept drawings with estimates of their impacts. Like, we could do it this way for this cost and this way, we could do it that way for this cost and that way, and so on.
Quite marvelous. Building the business case is another one. AI tools help you analyze data, seek and analyze market information, making predictions as to market size, sales, pricing, and cost, monitoring competitors’ activities.
There’s online software you can buy that monitors competitors’ activities, their launches, their pricing. There’s also software that’ll spot technological trends, technology assessment software, and you can get facts on a specific technical subject. For example, if you said, give me some information about auto driving automobiles, how they work, and whether the technology is up to speed, you’ll get that answer from ChatGPT.
There’s software products, for example, Hub is one of them. There’s a few others that analyze financial data, make revenue and profit projections, simulate different scenarios, gauge the impact of factors like price and competition. There’s even several software programs that I found that will write a business for you with prompts.
It’ll ask you questions and you feed it the information that it asks for and it’ll do the verbal part of the business case, the justification of the project, the risk assessment, etc. It’ll also do the full financial analysis with multiple scenarios. Risk assessment is a particular strength of AI.
So putting together a good business case using AI is sort of neat. Development and testing is an obvious place where AI has been used to a great extent. Two companies seem to be the leaders in the world here.
One is a company called Siemens, which is the biggest manufacturer, I believe, in Europe. Certainly the biggest exporter of physical goods in Europe. And the other one is GE, similar type of company.
Both make electrical engineering and mechanical engineering type products. AI tools can obviously create 3D models and generate technical drawings, create mockups online or virtual prototypes that can be tested, digital models, if you will. Some people call them digital twins.
It’s a twin online of the product. They can design products with the right features and dimensions just by verbal commands, design products that are more user friendly, more aesthetically pleasing. Again, all through verbal commands and or feeding information in.
They can develop digital models and virtual prototypes for rapid and iterative product testing. One of the most interesting examples I saw was of a small company in California called BYE Aerospace, BYE. Check it out online.
They’re developing an electrically powered small airplane. Obviously, it’s very tricky to design the airframe for such an airplane because it’s got to be lightweight. But it’s also got to be strong enough to withstand flight and weather conditions.
Normally in the course of a development, as the chief engineer said in an interview, that you get one or two or three iterations of the airframe, because it is very, very difficult to do simulations. You design an airframe, then it takes you two or three weeks to do simulations to all the various computer models. With this new approach, actually using software that is for sale from Siemens, that German firm I mentioned, they’re able to do an iteration every week.
And as a result, they can get closer and closer and closer to the perfect design. They figured they can get 98% close. And that’s remarkable by doing multiple iterations of this airframe to get it the right weight and the right strength.
AI will also do structural optimization to reduce the weight and cost of a product. It’ll optimize the design with a digital twin. In other words, there’ll be a computer model of the product and also a physical twin of that model.
And so the product can be tested in the field. Let’s say it was a tractor. You would have a prototype tractor in the field with sensors all over it.
And on your screen, you’d have the tractor virtual version. And you can monitor the tractor and collect data on its operation and basically do a really, really good performance assessment, much better than traditional field trials. AI can also be marvelous things that can create or discover products like chemicals and drugs.
A number of pharmaceutical companies and chemical companies are using what is called drug discovery or chemical discovery, using AI to predict the outcome of chemical reactions and thereby create entirely new products. For example, one Swiss company was desirous of developing pesticides that did not harm the environment. They used AI to do it.
Another thing people are doing is automating project management. If you were a project manager, keeping track of what’s going on, whether it’s late, on time, etc. I could go on.
One of the things I haven’t gotten into, and yet there’s a lot of material yet to talk about, is commercialization, production startup, market launch, optimized pricing. For example, Uber is using AI pricing to develop optimal pricing strategies. Advertising, Marcom.
And then finally, post-launch. Monitoring the product once the customer is using it. You know that every Tesla on the road has a digital twin back at headquarters or back in the engineering lab? So that people somewhere back in wherever that is can monitor how your car is performing, how your automobile is performing.
And a friend of mine who owns a Tesla says he gets messages periodically saying, you should watch out for this, or you should check that. Presumably automatically from the digital twin. Trouble is, folks, and this is going to be my last slide, the trouble is the adoption rate by industry in the West, I include Europe as well as North America, is not very good.
Despite the phenomenal results achieved by a handful of leading firms like GE or Siemens or Pfizer or Unilever or Procter & Gamble, the average firm is not doing nearly as well. And that is concerning. These leading firms have deep pockets, lots of money, and they have large IT departments, and they have a lot of experience about introducing new ways of working.
And they’ve done it right for the most part. This has come up from IBM studies as well. The early adopters got excellent results, and now they’re spending big bucks to convert the rest of their business.
The average company, however, is not doing so well. They don’t have as deep pockets, they don’t have the experience, and they don’t have the IT strength, perhaps, as some of these larger firms do. This is the adoption rate by activity by US and German companies combined.
The Germans had slightly higher rates. Ideation quite low. Market analysis was the most frequently used activity.
16% of firms that we looked at doing R&D, physical product firms, were using it to do the market analysis. 52% intend to adopt it, but intentions, nice words, but it hasn’t happened yet. Competitive analysis, low.
Making go-kill decisions. Managers are very unwilling to let go of the reins of decision making, even though current management decision making in product development is not very good. 70% of the projects that management approves for development end up being commercial failures or duds, or being canceled.
Their current decision making ability is worse than the toss of a coin, and yet they’re unwilling to relinquish the reins of decision making to AI. Product design, very, very low percentages. This concerns me as a citizen of North America, and should consider all of you, PhD students, or managers, or owners of small businesses.
China and India are substantially ahead, according to an IBM study done in late 2023, just released in 2024. Most concerning. The name of the game, guys, is this tsunami, this fourth industrial revolution, called artificial intelligence, combined with robotics, combined with biotech, those three technologies, but with AI leading the way, is coming on like a big wave.
The book, The Coming Wave, is a good read. You want to take a look at that. Like the first industrial revolution back in the 1700s, it is going to change the face of industry globally, and yet a lot of companies are not up to speed, and they’re going to get swamped.
My advice to each of you is to get up to speed yourself personally. At a conference recently, somebody asked the question of the speaker, who’s going to lose their job in this, in this new revolution, in this new transformation, when this wave hits? Who’s vulnerable? Whose jobs are going to be lost? And the speaker wisely looked at the person asking the question and said, the only people that are going to lose their jobs are the people that don’t know what AI is, nor how to use it. But if you are AI literate, you will, you will benefit, because you’ll be one of the few people in the room that knows how to run forward with it.
So that’s some advice here. Get yourself up to speed, your colleagues up to speed, your boss up to speed, and then get into an AI adoption program in product development for your own company. I mentioned product development specifically because the research has shown from McKinsey and others, from IDC research as well in Boston, that one of the biggest benefits of adoption of AI has not been improved cost structures in the production department.
Sure, that’s happened. But one of the biggest benefits has been improved innovation. That’s the number one, better innovation.
The other thing that I found, and this is good news for all of us, is that all the functions in the company, the only ones that are not, that do not lose jobs, is our research development and engineering, which is part and parcel of product development. So it’s an excellent area, big payoffs, not as much risks as other areas. Get into it and get into it now.
Author:
Robert G. Cooper
Professor Emeritus, McMaster University
https://experts.mcmaster.ca/display/cooperr
To learn more, read:
https://snyder.syracuse.edu/want-to-augment-a-products-aesthetic-design-use-machine-learning/