Data, not LLM
Think twice about the skillset of your data team
In my last post, I explored AI's boundaries in analytics. Today, I want to talk about a common but dangerous misconception: that data teams are naturally equipped for AI initiatives.
Nearly every data use case revolves around analytics, which requires clean, organized information. For years, data teams have honed their skills in wrangling spreadsheets, crafting SQL queries, and building dashboards.
But now, C-level executives demand, "We need to do AI!" And in the collective imagination, data equals AI. It seems logical, as data companies are quick to point the correlation between high-quality, governed data and successful AI projects.
This is where the misconception quicks in; while data teams master analytics and structured data (tables and columns that can be sliced and diced), the most intriguing AI use cases often involve unstructured data - processing natural language, analyzing images, or decoding audio files. This requires a very different skill set.
We're doing great things with AI for analytics, sure. It's an important subset of enterprise AI. But it's not the whole picture, and certainly not the focus of dedicated AI managers.
So why the confusion? There's a dangerous assumption that data teams can seamlessly transition to AI projects using unstructured data. The reality? Many data teams are still perfecting revenue calculations using structured data - and they are far from being able to build AI systems with unstructured data. This misalignment leads to frustration and missed opportunities.
As we explore AI use cases, we must recognize the distinction between data analytics and AI development. They overlap, but aren't interchangeable. We need to be thoughtful about team structures and resource allocation.
The path forward? Assess your data team's skills honestly. Consider separate AI teams with specialized skills. Foster collaboration between data and AI teams. Invest in upskilling if needed. And set realistic expectations with leadership.