Are you betting on AI? You must first consider product-market fit

The AI ​​boom not going to planOrganizations struggle to transform Investments in AI into reliable revenue streams. Companies are finding generative AI harder to deploy than they hoped. AI startups are overvalued and consumers are losing interest. Even McKinsey, after predicting $25.6 trillion of the economic benefits of AI, now admits that businesses need “organizational surgery“to fully exploit the value of technology.

But before rushing to rebuild their organizations, leaders should go back to basics. With AI, as with everything else, creating value starts with product-market fit: understanding the demand you’re trying to address and making sure you’re using the right tools for the job.

If you’re nailing things together, a hammer is fine; if you’re cooking pancakes, a hammer is useless, messy, and destructive. In today’s AI landscape, however, All gets hammered. CES 2024Attendees were amazed by AI toothbrushes, AI dog collars, AI shoes and AI products bird feeders. Even your computer mouse now has an AI buttonIn the business world, 97% of executives say they expect Generation 2 AI to add value to their businesses, and three-quarters of them are outsourcing customer interactions to chatbots.

The rush to apply AI to every conceivable problem leads to many products that are only marginally useful, and some that are downright destructive. A government chatbot, for example, told New York business owners to fire workers who complained of harassment. Turbotax and HR Block, meanwhile, have put online bots that gave bad advice as often as half the time.

The problem isn’t that our AI tools aren’t powerful enough, or that our organizations aren’t up to the challenge. It’s that we’re using hammers to fry pancakes. To get real value from AI, we need to start by refocusing our energies on the problems we’re trying to solve.

Furby’s Mistake

Unlike past technology trends, AI has the unique ability to bypass companies’ existing processes to establish product-market fit. When we use a tool like ChatGPTIt is easy to be reassured by his human appearance and assume that he has a human understanding of our needs.

It’s analogous to what we might call the Furby fallacy. When talking toys first hit the market in the early 2000s, many people, including some intelligence officials — we assumed that Furbys learned from their users. In fact, the toys were simply performing pre-programmed behavioral changes; our instinct to anthropomorphize Furbys led us to overestimate their sophistication.

Similarly, it is easy to misattribute intuition and imagination to AI Models —and when it seems like an AI tool understands us, it’s easy to miss the difficult task of clearly articulating our goals and needs. Computer scientists have struggled with this challenge, known as the “alignment problem,” for decades: The more sophisticated AI models get, the harder it is to give instructions with sufficient precision—and the greater the potential consequences of failure. (Carelessly ask a powerful enough AI system to maximize strawberry production, and it could turn the world into a large strawberry farm.)

Beyond the risk of an AI apocalypse, the alignment problem makes product-market fit for AI applications more important. We must resist the temptation to fudge the details and assume that models will figure things out on their own: only by articulating our needs up front and rigorously organizing the design and engineering processes around those needs can we create AI tools that deliver real value.

Back to the sources

With AI systems failing to find their own way to market fit, it’s up to us as leaders and technologists to meet our customers’ needs. That involves following four key steps, some familiar from Business 101 courses and others specific to the challenges of AI development.

1. Understand the problem. This is where most companies go wrong, because they assume that their main problem is the lack of AI. This leads them to conclude that “adding AI” is a solution in its own right, while ignoring the real needs of the end user. Only by clearly articulating the problem without reference to AI can you determine whether AI is a useful solution or what types of AI might be appropriate for your use case.
2. Define product success. It is essential to discover and define what will make your solution effective when working with AI, as there are always trade-offs. For example, one question might be whether to prioritize fluidity or precision. An insurance company creating an actuarial tool might not want a slick chatbot that failures in mathfor example, while a design team using generative AI for brainstorming might prefer a more creative tool even if it sometimes produces nonsense.
3. Choose your technology. Once you’ve defined your goal, work with your engineers, designers, and other partners to determine how to get there. You can consider different AI tools, from generative AI models to machine learning (ML) frameworks, and identify the data you’ll use, applicable regulations, and reputational risks. It’s critical to answer these questions early in the process: it’s better to build with constraints in mind than to try to solve them after you’ve launched the product.
4. Test (and retest) your solution. Now, and only now, can you start designing your product. Too many companies rush through this stage, building AI tools before they truly understand how they will be used. Inevitably, they end up looking for problems to solve and struggling with technical, design, legal, and other challenges that they should have considered earlier. Prioritizing product-market fit from the start helps avoid such missteps and enables an iterative process of progress toward solving real problems and creating real value.

Because AI looks like magicIt’s tempting to assume that deploying an AI application in any environment will create value. This leads organizations to “innovate” by firing off bursts of arrows and drawing targets around where they land. A handful of these arrows will actually land in useful places, but the vast majority will provide little value to businesses or end users.

To harness the enormous potential of AI, we must first aim for the target and then work hard to reach it. For some use cases, this may mean developing solutions that do not involve AI; in others, it may mean using simpler, smaller, or less attractive AI deployments.

No matter what type of AI product you’re building, one thing remains constant: establishing product-market fit and creating technologies that meet the real wants and needs of your customers is the only way to create value. Companies that do this will emerge as winners in the AI ​​era.

Ellie Graeden is a Partner and Data Scientist at Luminos.Law and research professor at the Massive Data Institute at Georgetown University.

Mr. Alejandra Parra-Orlandoni is the founder of Spire Tech.