The past few years we together have witnessed a surge in Artificial Intelligence that rivals the impact of the internet, email, and even the personal computer on consumers. Large tech companies are offering innovative AI-powered services, startups are blossoming, and hardware companies have seen exponential grow. AI is undeniably here to stay, except if all of this were an illusion?
I have spent a decade in the “so-called” AI field, I’ve seen significant shifts. Before Transformers, focus wasn’t on general intelligence, but on justifying AI use cases. Choices were deliberate, and operational costs mattered. It was a time when machine learning, though innovative, was carefully considered for tackling truly challenging problems.
My concern is that companies solely focused on building AI products might face rapid decline. Why? Because they often violate a fundamental principle of product leadership: solve a problem, don’t force-fit a technology. This applies even to multi-billion dollar companies.
Let’s use an analogy. Imagine using a complex platform like Apache Spark (a Big Data processing tool) for a simple math operation over a significant but reasonable dataset. Where a basic SQL query on a traditional database would suffice. It might not sound glamorous, but it avoids technical debt and minimizes operational costs — the true chain of value.
As engineers, our core role is to solve problems technically, optimizing processes, value chains, and business scalability. AI excels when traditional methods become unwieldy due to complexity.
A prime example comes from the long story of a client in the consumer goods industry. Their initial success came from understanding local customer preferences through direct interaction. As they expanded, replicating this approach became inefficient. Their vast product array led to significant food waste due to a disconnect from customer needs.
To re-gain a deeper understanding of customer preferences, we identified key factors influencing their decisions: brand loyalty, price sensitivity, search relevance, and purchasing habits. Because of the multi cultural scale complexity, we decided to build a suite of machine learning models that analyzed customer data, uncovering previously unseen patterns and trends. This allowed us to optimize product offerings without the need to understanding every customers individually at human scale.
This story underscores a crucial point: the store’s initial success wasn’t due to complex algorithms, but to a deep understanding of their local environment. AI strives to replicate this human ability to grasp complex situations, but on a massive scale.
While AI has sparked debates about its capabilities, it’s clear it functions best as a tool. Just like any tool, its effectiveness hinges on understanding the problem at hand. As the local store example illustrates, a deep grasp of the situation is crucial before reaching for a technological solution.
AI development and implementation can be expensive, both technically and in terms of human expertise. However, the true question lies in the return on investment. When used strategically to address well-defined problems that traditional methods struggle to handle at scale, AI can deliver significant benefits.
This includes generative AI, which can automate really complex tasks that are too time-consuming, especially when the problem is well-defined. Microsoft exemplifies this concept brilliantly with the term “Copilot”. It’s not just another chat tool; it’s from the ground up designed to address the core issue of productivity by acting as a helpful assistant to their existing software suites.
They may have also largely contributed to the current hellish corporate complexity themselves, but that’s a discussion for another time.