Strategic use of LLMs: innovation without complexity

In this blog, we examine how LLMs can drive innovation, but also how they can sometimes unintentionally add complexity. The key takeaway? Strategic use of AI can lead to impressive benefits, but only when applied in a smart and manageable way.

LLMs are versatile: they can summarize texts, answer customer queries, and even write code. For many organizations, they are a game-changer in the world of automation, offering benefits like cost savings and improved customer experience. Yet, there's a trap: not every problem requires an LLM solution. Trying to apply LLMs to everything often backfires.

Take, for instance, the rapid rise of chatbots in customer service. Many organizations eagerly adopted them, only to find that most customers still preferred speaking to a human. A study by Forrester even revealed that 54% of customers were dissatisfied with chatbots, primarily because they struggled to handle complex or sensitive inquiries. The lesson here? The latest technology is not always the best solution.

LLMs are especially valuable in the early stages of innovation. They enable you to quickly build prototypes, test the feasibility of ideas, and discover their true value. This allows you to experiment with new concepts faster than ever, without needing to heavily invest in their development.

Does an idea work? Then the next step is to simplify and refine the solution. LLMs help you find value quickly, but the ultimate implementation often revolves around a targeted, efficient approach.

In business process automation, the 80/20 rule often applies: the first 80% of a process is relatively easy to automate, but the final 20%—the edge cases, exceptions, and complex details—can be the most challenging. This is where LLMs can sometimes fall short. Many organizations try to automate entire processes with an LLM, only to discover that this last 20% requires unexpected effort in terms of accuracy and maintenance.

LLMs deliver quick results: less manual work, faster responses, and greater efficiency. But what often gets overlooked are the long-term costs of using these models. Hosting an LLM comes with ongoing infrastructure costs, such as Kubernetes services, GPUs, and powerful cloud computing.

Balancing performance and costs is therefore crucial. LLMs can achieve remarkable results, but the resources required can quickly add up. It’s important to seek a balance when implementing this technology that not only provides powerful performance but is also affordable and scalable in the long run.

As your organization relies more on LLMs to handle tasks, there's a risk that valuable knowledge may be lost due to reduced direct involvement in processes. Over-reliance on AI can create a gap between decision-makers and the day-to-day realities of the organization. It's like a manager who delegates everything and eventually loses sight of what’s really happening on the shop floor.

I firmly believe that innovation thrives when organizations remain actively engaged in solving challenges, rather than blindly outsourcing tasks to AI.

LLMs are incredibly effective at producing convincing, human-like responses. But as powerful as they are, they are not infallible. In fact, LLMs often generate inaccurate or irrelevant information—a phenomenon known as "hallucination."

For instance, while writing this blog, some references generated by AI turned out to be completely incorrect. Without a critical eye, it’s easy to develop a false sense of confidence in AI-generated content. That's why it's essential to make validation and human oversight an integral part of any AI process.

LLMs offer many possibilities, but a focused, strategic approach is essential. Whether you’re testing new ideas, automating tasks, or boosting efficiency, it’s important to stay focused on simplicity and sustainability. By using LLMs for specific, high-value use cases, you can avoid making solutions unnecessarily complex or costly. It pays to critically evaluate the right use cases and continuously maintain a balance between performance and costs.

Would you like to discuss how to use LLMs effectively and strategically within your organization? Get in touch with us. We are happy to share our insights and explore the possibilities for optimal AI application in your situation.