The views expressed here are my own.

As a Data Scientist at Switch Automation, I spend much of my time working with data, integrating and building analytical solutions, and turning complex information into something people can use. With artificial intelligence advancing so quickly, one question is becoming increasingly difficult to ignore:

Is AI coming for my job?

The honest answer is both yes and no.

AI is certainly coming for parts of my job. But I do not believe it is coming for the entire role. More accurately, AI is changing what it means to be a data scientist and what will be expected from us in the future.

AI can already do a surprising amount

Today’s AI tools can write code, explain statistical concepts, generate queries, summarize datasets, provide visualizations, and help troubleshoot errors. Tasks that used to take several hours can now be done with a click of a button.

Back in the day, I used to “google” and “stackoverflow” everything. Now, it’s “chatgpt” or “codex”.

In my work, AI now works as my technical assistant. It can write code, improve it and at the same time, debug possible errors.

That is incredibly useful – but that doesn’t mean it can work on its own.

An AI-generated query may look correct but still use the wrong business definition. It can give an answer right after you hit send, heck even write a blog if you want, but AI will not guarantee the authenticity.

AI can tackle a complex mathematics challenge from a simple prompt and is very good at producing answers. But the results may not be client-ready or be treated as the source of truth.

Data science is not just about code

As broad as it is, on top of working for a company where everyone is an IC (individual contributor), I find my role both satisfying and challenging at the same time. Instead of being scared of AI, having it on my side is a big plus.

Here are some stories where AI has helped me:

• Client integration on our platform is not always; (1) fill the form, (2) upload, (3) done. As much as I want it to be like that, it has never been. Different clients mean different data integrations and delivery models, lucky if we work with an industry standardized format. Many times, we need to go back and forth just to make the connection work. Mind you this is just the connectivity.

• Dealing with historical data is one of my favorite tasks; and one I am scared of. AI makes it easier to consolidate everything for me to review, but that doesn’t mean automating data straight into our databases. There are often records which are weird, incorrect and missing. My role is sensitive to polluting the databases, especially with backfilling large timeseries datasets where we have to be perfectly aligned.

• Workspace or Platform Reports are where we offer solutions for clients to make use of our platform to see their data, both historical and live (there’s more we can offer! Check us out! Oh, wait ads are not paid lol). It is currently both our strength and bane; we can generate reports the client wants but currently we have to deliver it by ourselves. AI helps with coding and debugging, but it can't always deliver the finished report without my involvement.

• Python Script – nowadays working from scratch is considered “old school” and most of the people in our industry are now called “vibe coders” (I personally didn’t find the new nickname cool, but it is what it is). Creation is only one part of the whole; nobody requests something from an AI and just deploys it; everyone knows “it is not perfect” hence the need to scrutinize every aspect of the result to ensure it’s aligned with the desired output.

Knowing how to write code remains important. However, knowing what the code should accomplish—and recognizing when it does not accomplish it—is becoming even more important.

So, should data scientists be worried?

No. Instead, we should embrace it more. Life is about evolution, and the same goes for our roles. We will be free from the repetitive part of the work and can focus on working with the other side of our role, developing new innovations by leveraging the AI automation.

AI can do a lot of things, but it can’t function at its best without the right expertise and judgement. My job is to understand the problem, work with solutions and provide beneficial results. And with the help of AI, do this a better way.

Perhaps the right question was “How will I use AI to become a better data scientist?”

About the Author

Pau is a Data Scientist at Switch Automation. He handles data from integrations to customer-ready product solutions.