The AI Use Case Pandemic

Why the Use-Case-Driven Approach Gives Me Headaches and Sometimes Even Stomach Aches

May 02, 2024

#beyondAI

In the world of AI, not all that glitters is gold. This is a principle I often discuss with teams and companies considering a deeper dive into AI Product Management. The allure of use-case-driven AI projects is undeniable, but the real gold lies in a product-driven approach that aligns closely with strategic business goals.

AI projects should be more than technological feats, they are strategic business investments. Failing to recognize this can lead to spectacular tech demonstrations that ultimately fail to integrate into or enhance business processes.

🔥 There is a great opportunity for AI Product Managers.

If the preference for use-case-driven approaches continues unchecked, companies risk not only financial losses but also the erosion of internal trust and credibility.

And the statistics prove this. There are numerous failed AI projects on record.

🎓 So, dear aspiring AIPMs, find out how you can prevent businesses from falling into this trap.

This newsletter aims to shed light on why embracing a product-driven approach could steer AI initiatives toward more sustainable success and better alignment with business objectives.

You will discover:

Why is there so much buzz about AI use cases?
Why do AI Projects fail?
What is an AI Use Case?
Use Case Driven vs. Product Driven
Takeaway for AIPMs

Have fun 🎳

Why is there so much buzz about AI use cases? 🧐

It’s actually simple: AI companies, consultancies, and other agencies want to showcase the capabilities of their AI products or services by demonstrating AI use cases. They often want us to be inspired and, consequently, buy their products and services.

And they are quite successful at it.

But do you know what is happening? Companies and individuals perceive those use case demonstrations as universally applicable.

You might doubt this, wondering, “Surely, people can’t be that naive?”

They aren’t necessarily naive—it’s a more complex issue. I've observed this phenomenon firsthand and believe it stems from a natural tendency among those new to a field. When individuals or organizations encounter a field as complex and promising as AI, they may have a limited initial understanding. This can lead them to passively absorb information without critically engaging with it.

It simply seems logical to them that a specific AI use case could perform well because they recognize similar processes or manual tasks within their own company or daily activities. The connection appears so direct and relevant that the allure of the use case is hard to resist.

And do you know where this led us?

Why do AI Projects Fail?

Right into a common trap—many AI projects fail to move from promising pilots to successful, scaled solutions. The chart I've shared illustrates just how prevalent these challenges are, with production issues topping the list. A staggering 47% of projects struggle with integrating AI into existing business processes and applications, which is no small hurdle.

Source: Gartner (used here)

This issue is further complicated by internal resistance—management and politics can often slow down or completely derail AI initiatives. Only then does the first skill-related issue come up, where teams lack the necessary expertise to manage and scale AI solutions.

It's clear why so many projects falter.

When AI use cases are presented as one-size-fits-all solutions, it's tempting to believe they will seamlessly fit into any operation. However, the reality is far more complex. Each organization's needs and challenges are unique, and what works in a controlled demonstration might not work in diverse, real-world environments.

So, while the initial allure of AI use cases is strong, driven by their perceived applicability and the promise of easy success, the journey from a pilot to a fully integrated system is fraught with obstacles. It's a reminder that without proper planning, support, and skills, even the most promising AI projects can struggle to achieve their full potential.

So the key challenge lies in differentiating between the surface appeal of a use case and its practical, in-depth application to one's specific context.

And that’s a problem.

What is an AI Use Case?

But let’s also clarify first what we mean by an “AI Use Case.”

What is it? An AI use case is a scenario in which AI technology is applied to solve “potential” real-world problems or enhance existing processes. It's a specific example where AI can “potentially” add value, tailored to particular needs or challenges.

I emphasize “potentially” for a reason. We cannot imply that these needs or challenges, once solved, will directly reflect in some kind of monetary benefit.

Example: Consider an AI system designed for predictive maintenance in manufacturing. By analyzing data from equipment sensors, the AI predicts when a machine is likely to fail, allowing for maintenance before costly breakdowns occur.

A valid AI use case, right?

But the effectiveness of such a use case doesn't solely hinge on its ability to predict failures. Its value is also measured by the return on investment (ROI) it delivers, which includes reducing downtime and maintenance costs, extending equipment lifespan, and improving overall operational efficiency.

So, if this AI system is working absolutely well but fails to generate value due to poor integration with existing workflows, one might question the wisdom of investing in it at this time.

🌟 The core of this argument lies in understanding that the successful adoption of AI technology extends beyond its mere functionality.

So it seems integration is more than just a fancy word. Let’s understand why:

An AI system must mesh seamlessly with existing business processes to influence meaningful outcomes. When there is a disconnect, even the most advanced AI solutions can become redundant. For example, if predictive maintenance alerts generated by AI are not timely or effectively communicated to the maintenance team, the potential benefits—such as preventing downtime or extending equipment life—are unrealized. This misalignment can lead to frustration, underutilization of the AI system, and ultimately, a poor return on investment.

Then we also have Employee Adoption and Usage. Effective integration also influences how well employees adopt and utilize the new technology. If an AI system is perceived as cumbersome or disruptive to daily tasks, employee resistance to using the system can undercut its potential benefits. Successful integration involves redesigning workflows in ways that enhance, rather than complicate, employee efforts.

🌟 I sometimes even have the feeling that, with increased AI maturity within companies and more AI products readily available for purchase, the real challenges we, as AI Product Managers face, are shifting from a technical to a more strategic and organizational focus. Not that it has ever been purely technical or devoid of strategic and organizational considerations, but there has been a noticeable shift in how I allocate my time. Whereas once, about 70% of my efforts were dedicated to building AI solutions and organizing development teams, it is now increasingly balanced, edging closer to a 50/50 split between technical oversight and strategic, organizational duties. It's becoming less about whether the AI can perform a certain task, and more about how it integrates into the broader business ecosystem and aligns with strategic goals.

Anyway, this is just a side note 🤓

Use Case Driven vs. Product Driven

So, is it always bad to be use-case-driven? "Build it and they will come" the old adage promised. The crux of being use-case-driven is that it often results in recognizing failure only after the solution is built; whereas, a product-driven approach anticipates potential pitfalls right from the start.

Starting with a use-case-driven approach is akin to shooting arrows in the dark, hoping to hit the bullseye. It's a scattergun method that lacks precision and foresight.

Being use-case-driven can be beneficial when it serves to inspire new applications within your organization or for your end customer.

Ultimately, though, it is essential to be product-driven.

Let’s understand the differences. The fundamental distinction between use-case-driven and product-driven approaches is rooted in their initial focus and how they navigate through the product development lifecycle.

Use-Case Driven:

Focus 🔍: Begins by identifying a specific application or capability of the technology.
Process ⚙️: Asks, "What can AI do?" This approach searches for potential applications of the technology, often overlooking whether these applications solve actual problems.
Outcome 🚧: This method can lead to technologically advanced yet essentially misguided solutions in search of a problem. Such products frequently fail to meet significant market needs, lacking a clear target audience or market fit.

Product-Driven Approach:

Focus 🔍: Starts with a clear understanding of a market need or a problem that needs solving.
Process ⚙️: Questions, "What do customers need, and how can AI solve it?" This ensures the technology is not just innovative but also purposeful and relevant.
Outcome 🏆: Involves rigorous upfront research, including user testing and market analysis, to confirm the market's demand before fully developing the product. The result is a product that aligns well with market demands and user expectations, facilitating better adoption and success.

Takeaway for AI Product Managers

To ensure that your company does not fall into the same pitfalls and become overly influenced by the advertisements of large AI companies, it's crucial to conduct a detailed analysis that goes beyond initial impressions:

Validation Against Specific Needs: Ensure the use case aligns precisely with the problem you are trying to solve. This involves understanding not just the end goal but the nuances of the processes that lead to it.
Feasibility and Integration Assessment: Consider whether the existing infrastructure supports the integration of the new AI solution. Additionally, assess the readiness of the team to adopt new technologies and adjust workflows accordingly.
```
🔥 I highly recommend at this point to also consider visualizing the entire business process, which should be enhanced or improved by AI. Use tools for Business Process Modeling. Here is a great video to explain what BPM is 
```

If you've understood why companies' use-case-driven approach gives me headaches and sometimes even stomach aches, then my first goal with this issue is fulfilled. Furthermore, if you've also grasped what your role as an AI Product Manager (AIPM) entails and how you can ensure your company doesn't fall into the misconception that being use-case-driven is the way to go, then: Mission accomplished.

However, if this isn't the case, don’t worry—there’s always room for a rerun. Think of it as binge-watching your favorite sitcom: the more episodes you watch, the better you understand the jokes. So, let’s rewind, play it again, and maybe this time, it’ll click.

And in the worst case? Well, we might need to consider enrolling you in a Use Case Rehab Program, complete with daily affirmations of "I will think product-first!" to get us all back on track. 😄

JBK 🕊

P.S. If you want to support me, for now, I am not accepting payments ☺️ However, you are welcome to share, like, and comment here or on my LinkedIn posts. This really helps me reach more people who are on the same journey as we are, and your feedback is invaluable in helping me improve the content iteratively ❤️.