Cursor for Analytics: Where it Fails and Where it Thrives
A step-by-step workflow walk-through with code samples and screenshots
If you’re an analytics engineer who wants to be the most productive, valuable version of yourself, you need to start integrating AI into your workflows. Whether you disagree with the ethics behind it or think it’s going to lead to brain rot, you need to start learning how to use it now.
My employer recently began mandating AI training for its engineers because, not only does it want the company to be ahead of the curve, but it also wants its engineers to keep up with the times. In just a few years, if you don’t know how to use AI, you probably won’t get hired.
I don’t say this to cause fear, but more so to motivate you to start learning these tools now. The sooner you learn them, the more of a leg up you have over those who wait until it’s absolutely necessary.
I also want to mention- it’s never been an EASIER time to break into analytics engineering or data. AI can help accelerate your learning beyond what would be possible without it. The knowledge gap between those hungry to learn and those who have been in the field for 20 years has never been faster to traverse.
Now that I’m done my pep talk on WHY you need to be using AI, let’s get into testing Cursor. If you aren’t familiar, Cursor is an AI-powered IDE specifically for software engineers. It has an AI agent built directly into the IDE, giving it access to your entire code base and terminal.
While it’s superpower is software engineering, I wanted to test it out to see where it thrives and where it fails for data and analytics work, specifically within a dbt project.
With learning to use these tools comes learning WHAT to hand over to them and what to keep tight reins on yourself. AI agents still get a lot of things wrong, so it’s important that you take the time to understand their strengths and weaknesses so you don’t become entirely dependent on them but rather leverage them for their optimizations.
In this article, there are 5 major analytics engineering tasks I use to evaluate Cursor’s effectiveness for dbt, SQL, and data model design work.
Test #1: Can it accurately make small code changes?
Test #2: How well does it create templates and document data sources?
Test #3: How well can it debug?
Test #4: How well can it write and optimize SQL code within my dbt models?
Test #5: Can it plan a data model design given a set of business requirements?
Let’s look at each of these in action and use the results to effectively measure the helpfulness of adding an AI IDE like Cursor to our analytics engineering workflow.
