Every data team will need to hire an LLM. Here's what you need to know.
Analytics has come a long way from its early days of spreadsheets and static reports. Over the years, business intelligence (BI) tools have been instrumental in shaping how organizations access and interpret their data. They have served us well, making data more accessible to the business, but for all their progress, a key challenge still remains: self-serve analytics.
BI tools promised self-serve — a world where anyone could ask and answer their own data-driven questions to inform their strategy — and yet, years later, we find ourselves no closer to it. In some ways, we’ve drifted even further away.
While businesses have evolved how they use data in the last few decades, BI tools have struggled to keep up. Teams are navigating ever-increasing volumes of information — leading to an uptick in questions that require more nuanced, dynamic, and iterative analysis than ever before.
But the linear frameworks of legacy BI dashboards weren’t built for that. While they’re great for answering a narrow scope of predefined questions like: “How many units did we sell?” they fail when it comes to answering the questions that clarify business strategy, like:
“What impact has our pricing change had on sales?”
“What is the likelihood for a customer to churn based on product usage patterns?”
These questions cannot be answered with one-and-done queries. They require exploration, iteration, and context.
This lack of self-serve access has created a natural divide between technical and non-technical users.
How many times have you looked at a dashboard, only to find yourself asking follow-up questions it couldn’t answer? Too often, stakeholders must rely on others to dig deeper — turning every new question into a request.
These requests land on the data team, who turn to notebooks and IDEs for Python and SQL-powered analysis. This creates bottlenecks, hinders collaboration, and leads to tool sprawl. Legacy BI tools and notebooks each have their barriers — one demands power-user expertise, the other requires SQL and Python knowledge.
We have, with no ill-intent, segregated our stakeholders into two groups: technical, and non-technical.
Technically advanced users: can navigate complex technical tooling and programming languages, but are burdened with constant requests from their less technical counterparts.
Non-technical users: rely heavily on pre-created dashboards but are limited by what’s already built for them.
In practice, our stakeholders fall somewhere on a spectrum that spans from highly technical ML engineers to no-code C-Suite, with the majority being somewhere in the middle.
The divide that existing data tools create underscores a fundamental problem in analytics: the need to bridge the gap between technical and non-technical users. This is the promise of self-serve and is a challenge that tools of the past have not solved. There is still a disconnect. Will AI fix it?
We would be remiss to not mention LLMs at this point. They are the answer to self-serve analytics that we have all been waiting for, right?
Wrong. Let’s be clear — LLMs don’t fix the problem outright. They are not a magic bullet. Instead, they are a crucial piece of the puzzle.
LLMs can help unlock accessibility for non-technical users and provide a new way of interacting with data, but they do not directly address the gap between technical and non-technical users.
The platforms that people do their data work in should allow creators of data products to produce data apps and dashboards without having to cater to a particular type of stakeholder.
The tool itself should allow the end user the ability to interact with many different modalities, be it SQL or Python, a chat-like experience with a LLM, or good old fashioned drag-and-drop exploration.
We need solutions that serve both data practitioners and business stakeholders, only then can we begin to move closer to the promise of an analytics tool that truly works for everyone.
We’re building Hex to be this platform where anyone, regardless of skillset, can work with data in a unified AI-powered platform — and LLMs are a key part of this vision, helping bridge the gap between technical and non-technical users. They enable more natural interaction with data — whether through direct querying, summarization, or intelligent assistance and when paired with a platform like Hex, they can make data exploration even more seamless.
Let's demystify them a bit, shall we?
Read on in our eBook.