Beyond the LLM Hype: Revitalizing Stalled AI Projects for Genuine Business Value

The Real World of AI: Experiences from Indesia's Datathon

Last week, I had the privilege of advising at Indesia's second Datathon, an event fostering collaboration between industrial companies and tech providers. The focus was on identifying real-world AI solutions. The outcomes were staggering, with use cases having the potential to impact businesses to the tune of tens of millions of Euros, as well as achieving sustainability targets.

You might imagine this was achieved through advanced AI or complex models like Large Language Models (LLMs). Yet, the solutions came from teams using straightforward machine-learning models for regression, classification, and prediction. These teams, made up of domain experts, data scientists, and engineers, were able to address challenges that had puzzled organizations for months or even years.

Moving from Model to Business Performance

McKinsey's report, "Notes from the AI Frontier," supports this reality, concluding that most ML applications enhancing existing business operations are direct and straightforward. Deep learning's power to draw patterns and make predictions can lead to significant operational enhancements. This underscores the need for businesses to pivot from focusing on model performance to emphasizing business performance.

This shift doesn't undermine the role of models in prediction but prompts us to translate those predictions into tangible business value. A high-performing model that doesn't lead to action yields no business value. Therefore, AI deployment success lies not only in crafting advanced models but in integrating them into core business processes and workflows.

Reassessing Business Impact and Addressing the Barriers

The accelerated problem-solving observed at the Datathon can be attributed to factors like advanced data sharing with the teams and employing a unified, user-friendly tech platform. The platform offered a framework for data governance and access, promoting collaboration within the team. This enabled the team to use tools and libraries that best suited their needs without impeding their focused work.

What stood out was the holistic approach to solution delivery and business impact assessment. The evaluation wasn't limited to profit increases or cost savings. It also considered the company's risk mitigation strategies and the initiatives' impact on society and the environment, highlighting sustainability.

These insights align with a Gartner survey that found 46% of AI projects don't reach implementation due to difficulties in understanding and quantifying the business value of AI. The Datathon's methodology addresses this by prioritizing business performance over model performance during evaluation.

Takeaways: Pivoting Beyond the Hype to Achieve Real Business Value

So, what's the takeaway for businesses? The real value of AI lies not in its complexity but in its applicability. It's time to shift focus from the (deserved) LLM hype and revisit stalled AI projects, aligning them with holistic business values. Large Language Models like OpenAI's GPT-4 may be trending, but the true value of AI lies in its tangible business impact, including profit, savings, risk mitigation, and sustainability.

Don't get lost in the LLM hype but leverage it. If your team needs a proven methodology to value AI initiatives or assistance transforming stalled AI projects into holistic success stories, don't hesitate to reach out. Together, we can leverage AI (including LLMs) to create not just business value but also a sustainable and secure future. Contact me today to embark on this transformative journey.