IBM AI watsonx decision predictive engine

IBM Think 2024: Could AI Help Figure Out What Customers Will Want?

Every industry has the same problem: figuring out what customers will want in a three-, five-, or 10-year timeframe. The reason we need to know what customers will want is that it takes from three to 10 years to create a new product, evaluate it, assure it, then bring it to market. Apple’s late CEO Steve Jobs figured that out early on.

At IBM’s Think 2024, the Q&A started with IBM’s CEO, Dr. Arvind Krishna, who said that whether it is hardware, software, or some mix of services, IBM will package what the customer wants. I’m ex-IBM, and my final job there was in marketing. One of the products that crossed my desk that I refused was a product IBM had spent several years and $20M developing, but I couldn’t figure out why anyone would want it. After doing a $20K study, it was determined I was right. No one wanted the product. It was an answer that would have been far more useful before the $20M was spent.

Steve Jobs argued that customers don’t know what they want and successfully showcased that if you adequately fund marketing, you could convince customers they want something and thus successfully anticipate a need because you created it. But even Apple seems to have forgotten that now. You can predict an outcome if you work to create that outcome, but it seems like most vendors thought the movie Field of Dreams was a manual on how to do things because they don’t adequately fund marketing for new products.

I think IBM’s watsonx, specifically, and AI generally, could fix this problem by helping to define a process and necessary marketing budget to ensure success.

Creating a Better Predictive Engine

We often focus AI on tactical projects like telephone sales, autonomous vehicles, robots, medical diagnosis, creating pictures or documents, and answering tough questions. But the big efforts, like Earth 2, are attempting to become more predictive. In fact, the industry is starting to focus AI on predictive securities trading.

The reason so many products fail in the market isn’t always that they were badly matched to the customers, but more often the fact that they are poorly marketed. The reason they are poorly marketed is that marketing isn’t treated like a strategic part of success. Ideally, when the initial product plan is set, what it will cost to market it and what the campaign should look like should be built into the budget.

The reason this isn’t done early is that those promoting the potential new product want to get a ‘yes’, so they leave or under-budget some costs, and marketing is a major part of that effort. But that means the product isn’t adequately funded, and much like a race car that isn’t given enough gas, the result is often a failed launch.

Granted this budget will reduce the number of products the company could launch in a given year, but it would better ensure those that are launched are successful. You don’t get credit for the most failed or marginally performing products. You get credit for successes.

Examples of this were Windows 95 and the Xbox, which were adequately marketed, while Zune, Cortana, and especially the Windows Phone were undermarketed and failed.

Training a Better watsonx Decision Engine

I’m focused on watsonx because it isn’t just the most mature and trusted AI platform. IBM has a century of sales data that could be used to train it so that it provides a much more accurate decision matrix. It could then lay out not only what the product should contain but the best path to market. It would outline a budget that included an estimate of what it would cost to successfully market the result, who the target customers are, and predict what feature set should be prioritized for the world that will exist when the product is ready to go.

And if the cost was beyond what the company would be willing to pay, then the product is killed early on rather than consuming resources and then failing, having a cascade effect on other products that, in turn, would also be inadequately funded.

Wrapping Up

One of the other jobs I had at IBM was as a Competitive Analyst. I discovered that executives don’t like to be told their baby is ugly. Before we were disbanded, we pointed out repetitively that the financials behind a proposed new product would bankrupt the division. Executive management was tired of that negativity, disbanded the team, brought out the product, and bankrupted the division.

Part of this problem is called Argumentative Theory which argues executives feel the need to seem right, especially if they are wrong, and look at those that provide negative information regarding their decision as a rival.

AI isn’t another person, and actually being wrong can be career-limiting, so packing a watsonx instance as an executive decision support tool could not only provide better information from which to make a better decision but do so without seeming as if it was challenging the executive’s authority.

Interestingly, I think IBM could use this tool itself. In the past, IBM has had a problem with top executives being left in the dark on critical decisions, causing those decisions to go badly. John Akers, the only fired IBM CEO, failed because he was isolated by staff from reality. Assuring that never happens again would be a valid goal for a future version of watsonx. Helping CEOs see what customers will want in the future, allowing them to see the full cost to assure a successful launch, and giving them the opportunity to better choose between potential new offerings based on total costs as well as projected revenues would help to avoid catastrophic outcomes from incomplete product budgeting and inadequate future customer focus.

Scroll to Top