AI enables “talking to our data” – really?

AI enables “talking to our data” – really?

You are probably getting tired of all this AI talk, but bear with me. Whether you are working with a lot of data day in and out or a product leader working with data drive products, finding answers to your data driven questions in the most effective manner and finding new, better and easier ways of interacting with your data is likely something on your mind.

Historically we have relied on spreadsheet jockeys (hoping their source data was accurate), analytics tools, database experts and data analysts to help us extract relevant the answers from our data. But generative AI is starting to offer us a new, and exciting way to interact with our data using natural language, doing it at our own pace. Those of you who are using generative AI tools already probably already know the data analyst tool that ChatGPT offer us as one of the many tool out there. But these tools still require a decent level of understanding of the data being processed. But what if you could just ask a simple question, get the answer, and move on?

Well, these solutions are upon us. Not broadly implemented yet, not ready for wide use in production for every use case, but here and ready to use now. In layman’s terms, what these systems allow us to do is to use a natural language interface combined with your subject matter expertise to interact with multiple relevant data sources such as databases and getting answers. It is like talking to your data….

So if these things exist, why aren’t we all using them now you may ask? Well, for starters, this technology is still relatively new. Secondly, these systems need to be trained to understand your domain and your data in order for them to provide meaningful responses. Thirdly, deploying AI systems in a larger scale environment means answering questions on how to deal with personally identifiable information (PII), ensuring reliability and scalability, ensuring the systems provide correct answers in all use cases (generative AI systems are still prone to “hallucination” in cases), and that users know how to effectively interact with these systems, to name just a few examples. These are all challenges that can overcome, but there isn’t a quick fix. On the other hand, there are applications for this technology in search and analytics that are ready now.

The key takeaway is this: these emerging technologies are here, they will stay, and they will change the way we interact with our data. As a minimum you should be aware of the existence of these tools. If you can, start learning about them to understand what they can do for you!