Cutting Through the AI Noise
If you’ve used AI in any capacity, you quickly realize its potential and the direction in which individuals and businesses must pivot to adopt these tools into their systems. I use AI as an executive assistant, a research partner, a sounding board, and my first stop for information. Personally, I rely on Google Gemini; my primary client recently rolled out Gemini enterprise-wide, encouraging all staff to integrate it into their daily routines.
AI is a powerful tool, and it is only going to get better.
Naturally, I want to know how AI can integrate with data analytics systems. We can envision opening a screen and saying, ‘Show me last quarter’s sales figures, broken out by regional sales manager, as a vertical bar chart with stacked profit margins,’ and—voilà—there it is. But how do I make that happen? How do I bridge the gap between data sitting in a legacy database and a language-driven AI model that can accurately analyze data on the fly?
I am currently architecting and testing local AI environments to bridge this gap. My goal is to find what actually works—and what doesn’t—so I can bring grounded, real-world solutions to my clients.
The two main components used to build this bridge are Retrieval-Augmented Generation (RAG) and Low-Rank Adaptation (LoRA) training. The key is accuracy. The last thing a leader wants is a model hallucinating and providing skewed data used to make critical decisions.
As I build prototypes in my local environment, my thoughts always circle back to feasibility. There is a ‘coolness factor’ in an AI-driven analytics platform, but does it make the best business sense regarding investment, computing requirements, and risk? If a company has already invested heavily in a BI platform, does it make sense to convert it, or is a hybrid approach better—where specific data is extracted from the BI platform and fed into an AI model for specialized analysis?
I also foresee a near future where tools like IBM Cognos, Oracle Analytics, Qlik, and PowerBI have integrated AI agents that allow natural language to interact directly with the analytic engine.
As I continue to learn, I’ll share more. In the meantime, feel free to reach out with your thoughts—I’d love to hear from you.