Should we stop outsourcing our analytical expertise to LLMs, or is this just the inevitable future?

Let’s look at this through the lens of conversational analytics—the ability to talk to your data in plain text and receive an answer without analyzing your graphs, writing a single line of SQL, or building a single dashboard.

By now, Google has integrated conversational analytics into most of its analytical tools. Google Analytics 4’s standardized analytical question/answer functionality was fortunately replaced by a Gemini-powered AI advisor. The old functionality was too basic to get any meaningful insight if you have the analytical ability to interpret a table or graph. For reference, it would answer questions like: how many users visited your website last week?

The AI advisor is a clear improvement in terms of depth of insight. It spots trends in your data and draws conclusions. Exactly the type of work that shapes the field of analytics. And let’s be clear—the conclusions are not half bad.

A clear improvement, right? Yes and no. Outsourcing more and more analyses to an algorithm will ultimately erode your ability to do the analysis yourself, and to verify whether the analysis is correct. It creates distance to the performance of your website and the skills that go into spotting trends and drawing conclusions. Someone without analytical experience won’t know when the machine is right or wrong.

I see the same pattern emerging in conversational analytics for BigQuery. Here, the agent surfaces its reasoning and the generated SQL behind every answer, then synthesizes the insights into a concise summary explaining the “why” behind the numbers.

Again, there’s a paradox: in order to verify the results, you still need SQL experience. If you outsource the practice of writing SQL, how are you meant to verify the results?

You are not removing the need for analytical skills; you are just moving them one step downstream. In organizations where decisions flow from data, misinterpretation at scale creates compounding risk. That’s why this trade-off matters.

Every time the agent writes the query for you, you lose a small opportunity to sharpen your own understanding of the data. For organizations where analytics maturity matters, that is a trade-off worth thinking about carefully. Because the field of analytics is not dead yet. Algorithms are not yet perfect.

You still have to go looking for insights, whether it’s in GA4, BigQuery, or any other analytical tool. Once meaningful insights come to you without any errors or mistakes, that’s when the field of analytics will be truly dead.

Conversational analytics is not the future of analytics. It is a feature of it. The future still requires people who know what they are doing.