Betting on AI? You need to consider product-market fit first

AI Boom Things don’t go as plannedOrganizations are struggling to turn their AI investments into a solid revenue stream. Companies are finding generative AI harder to adopt than they hoped. AI startups are overhyped and consumers are losing interest. Even McKinsey $25.6 trillion Regarding the economic benefits of AI, companies are currentlyTissue surgery” allows you to maximize the value of your technology.

But before rushing to reinvent their organizations, leaders need to get back to basics: With AI, as with everything else, value creation starts with product-market fit: understanding the demand you’re trying to meet and making sure you’re using the right tool for the task.

A hammer is useful for driving in nails, but it is useless, dirty, and destructive for baking pancakes. But in today’s world of AI, all We are suffering a blow. CES 2024Participants were amazed by the AI ​​toothbrush, AI dog collar, AI shoes, and more. Bird feeder. Computer mouse too There is an AI buttonIn the business world, 97% of executives Respondents expect the AI ​​generation to add value to their business, with three-quarters turning to chatbots to handle customer interactions.

The attempt to apply AI to every conceivable problem has resulted in a number of products that are either barely useful or outright disruptive. Government chatbots, for example, Speaking to New York business owners There’s even a bot that will fire employees who report harassment. Meanwhile, Turbotax and HR Block are I gave bad advice About half as often.

The problem isn’t that AI tools aren’t powerful enough or that organizations can’t rise to the challenge; it’s that they’re using a hammer to fry pancakes. To get real value from AI, we need to start by refocusing our energy on the problems we’re trying to solve.

The Furby Fallacy

Unlike past tech trends, AI tends to short-circuit a company’s existing process for establishing product-market fit. When using a tool like ChatGPT, it’s easy to take comfort in its human-like nature and assume it understands your needs like a human would.

This is similar to the so-called Furby fallacy. When talking toys first appeared on the market in the early 2000s, many people, especially some Intelligence Agency Officials — We assumed Furby was learning from its users, when in reality the toy was simply executing pre-programmed changes in behavior. Our instinct to anthropomorphize Furby led us to overestimate its sophistication.

In the same way, it’s easy to mistakenly attribute intuition and imagination to AI models. And when we feel like our AI tools understand us, we tend to skip the hard work of clearly expressing our goals and needs. Computer scientists have been working on this challenge, known as the “alignment problem,” for decades. The more sophisticated an AI model becomes, the harder it becomes to issue instructions with sufficient precision, and the greater the consequences of failing to do so. (If you carelessly instruct a powerful enough AI system to maximize strawberry production, the world might just fail.) Big strawberry farm.

AI apocalypse risks aside, the alignment problem makes establishing product-market fit even more important in AI applications. You must resist the temptation to blur details and assume the model will figure things out on its own. Only by articulating your needs from the start and rigorously organizing your design and engineering process around those needs can you create AI tools that deliver real value.

Back to Basics

Because AI systems cannot find their own path to product-market fit, it is our responsibility as leaders and technologists to meet our customers’ needs. That means following four key steps, some familiar from Business 101 classes and some specific to the challenges of AI development.

1. Understand the problem. Most companies make mistakes here because they start from the premise that the primary problem is a lack of AI. This leads to the conclusion that “adding AI” is a solution in itself, but ignores the actual needs of the end user. Only by clearly stating the problem without mentioning AI can you figure out whether AI is a useful solution or what type of AI is appropriate for your use case.
2. Define the success of your product. When working with AI, there are always trade-offs, so it is important to discover and define what is needed to make the solution effective. For example: Fluency and accuracyInsurance companies developing actuarial tools Math FailureFor example, a design team that uses gen AI for brainstorming might prefer a more creative tool, even if it occasionally speaks gibberish.
3. Choose your technology. Once you have a clear goal, work with engineers, designers, and other partners to figure out how to get there. Consider different AI tools, from generational AI models to machine learning (ML) frameworks, and identify what data you’ll use and any relevant regulatory and reputational risks. It’s important to address questions like these early in the process; it’s better to build with constraints in mind than to deal with them after you’ve released your product.
4. Test (and retest) your solution. Now, and ever more, is the time when you can start building your product. Many companies rush to this stage, creating their AI tools before they actually understand how they will be used. Inevitably, they end up searching for problems to solve and wrestling with technical, design, legal, and other challenges that they should have considered sooner. Prioritizing product-market fit from the start avoids these missteps and enables an iterative process of progress to solving real problems and creating real value.

Because AI seems like magic, it’s tempting to think that any AI application deployed in any environment will create value. So organizations try to “innovate” by firing arrows at a rapid-fire rate and targeting where they land. Only a handful of those arrows land in places that are actually useful, while the vast majority create little value for either the business or the end user.

To realize the vast potential of AI, we must first picture the target and then do all we can to hit it. For some use cases, this might mean developing a solution that doesn’t use AI, and for others, it might mean using simpler, smaller-scale, or less-glamorous AI deployments.

However, one thing remains constant when building any AI product: the only way to create value is to establish product-market fit and create technology that meets real customer wants and needs. Companies that get this right will emerge as winners in the AI ​​era.

Ellie Graden: Luminos Law Firm Research Professor at Georgetown University’s Massive Data Institute.

M. Alejandra Parra Orlandoni Spiraretec.