Here's what a day in my life as a data scientist looks like:
Every morning, I wake up, make myself a cup of coffee, and open my laptop.
Then I let AI do my data science job as I browse Instagram and sip on iced coffee, while making YouTube videos like this one.
Sounds like a dream job, right?
Well… except that I lied.
That's not true.
AI hasn't automated my data science job just yet, despite what you might read on the Internet.
I've been seeing a lot of posts lately on how AI is going to replace data scientists in the next 5 years.
For a video version of this article, click here.
The World Economic Forum has also listed data science as one of the most at-risk jobs from AI, causing many aspiring data professionals to panic.
I believe that much of this panic and doomsaying stems from the oversimplification of data science roles.
A data science job is pretty difficult, and in this article, I'd like to explain exactly what we do on a daily basis.
This will help you understand why AI has only been able to automate a small portion of a data scientist's role — because there are complexities that require a strong understanding of math, logic, and reasoning (not AI's biggest strengths).
Article Chapters
I have divided this article into 3 chapters — here's what you'll learn in each of them:
- Section 1: What do data scientists do
- Section 2: My opinion on whether AI will take data science jobs
- Section 3: How to get a data science job that is secure from tech layoffs
1. What do data scientists ACTUALLY do?
Here's what a day in my life as a data scientist really looks like:
It's a relaxing Monday morning, and just as I reach for my first sip of coffee, our product manager storms in.
"Our new feature isn't performing well." He says, looking unhappy.
"And you need to figure out why."
After asking ChatGPT, Claude, and our internal AI platform, which all fail to figure out what the issue is, I finally put my coffee down.
Looks like I have to do this myself.
Here are some steps I'll take to figure out why our product might be underperforming:
- First, I look into user behavior data and do some preliminary analysis using Python and SQL.
- Then, I speak to engineering teams to understand how data collection is being done. I make sure that the data is being tracked accurately, and that all metric calculations and ETL pipelines are up to date.
- If I still can't find the issue, I create A/B tests and segment users into different groups, to understand if a specific group of users are causing this dip in performance.
Essentially, my job is to use data science techniques to build and improve a company's products.
If something goes wrong or a product is underperforming, I've got to figure out why.
Let's break this down further with an example.
What do data scientists do (Example)
Every year, Spotify releases an annual recap called Spotify Wrapped that has over 100 million users worldwide.
A product team works to build this recap.
So you'll have a product manager who oversees the entire feature, a designer who designs it and makes it look all pretty, an engineer who builds it out.
Looks simple, right?
Just count some songs?
But behind all those pretty animations, there's a product data scientist like me.
Here's what the data scientist does.
First, they have to figure out what to include in the recap.
What do you show first?
- The top artist
- Total minutes listened?
- Your music persona?
This might sound simple, but even a slight change in the way you structure this can lead to a huge loss in engagement and even a revenue loss.
In Spotify's case, the data scientist needs to decide things like:
- What user segments to build
- How to calculate user metrics
- The product's final structure.
They will also run experiments that test different versions of Spotify Wrapped on different user groups.
- Version A shows artists first.
- Version B starts with minutes listened.
- Version C tries a completely new animation.
Now, sure, AI can help with some of this.
It can write some SQL queries, but even that requires a lot of prompting by someone who really understands the data.
To some extent, it can help you make pretty visualizations, and even do ML and statistics.
But here's what AI cannot do.
AI cannot execute an entire data science lifecycle, which involves:
- Deciding what metrics to capture based on the product requirements and domain.
- Working with engineers to collect relevant data from the platform and built ETL pipelines to calculate the metric.
- Running A/B tests and monitor performance of the product, and turn this insight into a business recommendation.
- Analyzing data and communicating insights to stakeholders for next steps. For example, explaining to the product lead why Spotify should show the top artist screen first, before the minutes listened.
Apart from programming and ML, a good data scientist also has the following skills:
- Product knowledge: You've got to know your product like the back of your hand to be able to even define what metrics you need to measure. This means you have to read a lot. At times, I can spend an entire day just reading through product requirement documents and engineering specs.
- Engineering and math: You must learn to build ETL pipelines and database design, and have a strong grasp of mathematics.
- Collaboration and business impact: You have to work with design and strategy teams to really understand what the business needs. And you've got to deliver on this business requirement using scientific methods. This is the most crucial part of your job. It's why companies hire you.
At the end of the day, the company has one goal.
To make money.
Any data scientist who helps companies make key product decisions are extremely valuable.
Because these decisions are worth millions of dollars.
And actually driving business growth?
It isn't just achieved with a beautiful dashboard or programming knowledge, but a deep understanding of the company and its users.
Of course, data scientists who don't have these skills will be replaced — but that's because they aren't driving business growth — not due to AI.
2. Will AI Take Data Science Jobs (My Opinion)
I use AI everyday to automate my Python code and SQL scripts.
However, that's a very small part of what I do on a daily basis.
All of the brainstorming, metric calculations, statistical rigor and business understanding still comes from me — the data scientist.
And that's why AI doesn't worry me for now.
Because it doesn't have the logic and reasoning skills required to do a huge chunk of my job — the part that drives business growth and gets me paid.
If I can leverage it to help me become a better programmer or even automate some parts of my workflow, that just makes me a more efficient data scientist.
Which just means faster product releases in a shorter time-frame.
I also want to emphasize that as a data scientist:
Your worth increases exponentially with experience.
Most of the skills mentioned in this article (such as product knowledge and business expertise) are gained through experience, not online courses and bootcamps.
This is why entry-level data science positions are less common and more competitive, while you'll have an easier time finding jobs as a mid/senior level data scientist.
How to secure your data science job from layoffs?
Tech layoffs impacted around 60,000 jobs in 2024, and the trend is predicted to continue in 2025.
While this is concerning, there are measure you can take to secure your data science job from tech layoffs.
After speaking to colleagues who have worked in the field for over 7 years, the #1 advice I've received is to choose business-facing data science positions.
Let me explain.
Many data science roles are focused on future impact rather than immediate revenue growth.
I've worked as a data scientist whose core focus was on experimentation — trying out new things that might add value to the company in the future.
However, this is a risky position to be in.
In a role like this, you're a "nice-to-have" to the business, not a "must-have."
When companies decide to reduce their workforce, it's easier to let you go than someone who is actively driving revenue to the business.
However, if you choose a data science position that drives direct revenue impact — one in which your decisions bring immediate value to the company — your job will be considerably safer.
For example, if you work at Spotify and are able to advise the product team on which UI will bring in more revenue to the company, your job has a direct revenue impact.
This means that you're more critical to the business and are less likely to get replaced.
The job market is unpredictable right now, and there is a lot of doomsaying around data science jobs. As an aspiring data professional, this can feel demotivating.
I hope that this article managed to clear up the confusion surrounding whether data science roles will exist and helps position yourself better in today's competitive tech landscape.