Two Paths to AI Product Development Success

When you look at PC OEMs, you’ll notice there are two product paths they typically take. The most popular is to build a high variety of products, even though they can’t afford to market any one of them very well, with the hope that potential customers will find one of those many products ideal. The other path is mostly just followed by Apple which builds a far more limited line of products but uses marketing to convince buyers that those few they do build are ideal for them.

Looking at the financial results from both approaches, it is strangely ironic that Apple’s approach is by far the most profitable yet is the least used by other companies. This is largely because tech companies are run by engineers. Engineers build things but don’t understand marketing. Steve Jobs wasn’t an engineer. His skill was manipulation and it made him one of the most powerful marketers of his generation. This is the foundation of Apple’s current success.

But with AI, could you enhance both approaches? Which one would be better and would result in higher sales, higher margins and a higher valuation?

The AI-Enhanced Shotgun Approach

I worked in competitive analysis for a time. We would use surveys and focus groups to determine what future products should have that would be attractive to customers. However, there is a big problem with this approach as I learned when studying GM’s competitive analysis approach.

Back in the 1960s, GM and Ford were concerned about Japanese cars, so they both took those cars and analyzed them down to the nuts and bolts that held them together. That analysis resulted in the Pinto and Vega models, both of which were widely panned at the time for not being competitive. What happened?

It took five years to do the analysis and create the cars using that initial competitive assessment as the template. But in the course of those five years, Japanese cars advanced significantly so that when the Pinto and Vega came out, while they were better than the Japanese cars of the ‘60s, they weren’t any more competitive than those Japanese cars would have been in the 70’s.

To fix this, you’d need to extrapolate from that initial analysis while looking at the rate of improvement, anticipate the technology of the future, then set a far higher bar that targets the world of the future.

Properly trained AI could help with this as it can aggregate a variety of information sources, identify trends, and potentially project those trends into the future. In addition, 3D printing technology can vastly lower the time it takes to prototype new designs, and NVIDIA’s Omniverse can create simulations that show that design as it would appear against anticipated future competitors.

In addition, AI can improve manufacturing efficiency and reduce the time it takes to repurpose a manufacturing line, potentially cutting down the time it would take to develop and build a new product so that the result is closer to the time of the analysis.

AI can therefore be used to set a more realistic bar, can cut down the time it takes to build a prototype using 3D printing, and also cut down the time it takes to spin up a manufacturing line dedicated to the new product.

Still, because there are so many variables, our ability to predict what that future product should be will only be slightly enhanced, resulting in a remaining significant risk you’d guess wrong.

Targeted Approach

The targeted approach flips this process. You don’t do focus groups or surveys, you just look at the products that are selling well, project into the future what you think people will want to buy, and then create a result that you wrap with sufficient demand generation marketing to drive people to that more limited line.

(wavebreakmedia/Shutterstock)

AI is already capable of bypassing storyboards to create powerful AI-generated photo and video content that can be assessed by customers for effectiveness. AI personas can be created that can assess the scripts or AI-generated content to see which is the most compelling.

While AI can still be used as it is in the shotgun approach to create a product that can be more easily marketed, instead of creating a bracket of different products based on the AI answers, you instead have it help you pick the most likely configuration across the variety of potential customer AI personas and use the ad creation approach above to drive customers to the product you’ve specified.

You can even use services like Indiegogo to see if your initial concepts are hitting your audience effectively, and then recraft the product and messaging to optimize the result.

This is a far simpler approach. As Apple has demonstrated, it’s riskier if the marketing isn’t done right, but it is also potentially the most lucrative because there’s far less chance you’ll have to sell off or dispose of large numbers of unsold products if people don’t find some of them attractive.

It is also faster and easier to spin up a marketing program than it is to spin up a product, so this approach allows you to respond more quickly to changing market dynamics and emerging new competitive threats.

Wrapping Up:

Of the two approaches used by tech companies, the more common shotgun approach continues to be the least efficient and the least profitable. The targeted approach has both cost and timing advantages due to its reduced complexity. AI can improve both processes but will have a greater impact on the targeted approach because it is far easier to modify a marketing strategy than it is to modify a product line. But the lack of marketing experience and authority in tech companies still makes this approach a non-starter even though it is more performant. So, I expect AI will have little impact on the go-to-market policies of the firms, and that once Apple spins to use AI internally, the profitability and product success the company currently enjoys will pale in the face of its future performance.

Finally, one of the big problems we will need to solve with AI is people misusing the technology. Rather than using it to make a better decision, they’ll instead use it to validate decisions that have already been made, decisions that might not be good decisions at all. Fixing this last issue is problematic because it goes against existing practices in engineering-driven firms. If it isn’t fixed, AI could actually make things worse by speeding up the creation of products that sell poorly, thus incurring more cost than the slower time-to-market enjoyed by the human-driven alternative.

Again, this points to the need to use AI to improve the quality of decisions before you use it to speed up how quickly things are done, otherwise you’ll simply be making more bad decisions more quickly and that typically doesn’t end well.

About the author: As President and Principal Analyst of the Enderle Group, Rob Enderle provides regional and global companies with guidance in how to create credible dialogue with the market, target customer needs, create new business opportunities, anticipate technology changes, select vendors and products, and practice zero dollar marketing. For over 20 years Rob has worked for and with companies like Microsoft, HP, IBM, Dell, Toshiba, Gateway, Sony, USAA, Texas Instruments, AMD, Intel, Credit Suisse First Boston, ROLM, and Siemens.