Hello Reader,

I started my career exactly 10 years ago as a data analyst in healthcare.

I loved every bit of it.

The puzzle-solving. Digging into a messy dataset and slowly making sense of it. Writing code to clean and transform data exactly the way I needed. Building a visualisation and watching the story emerge.

There was something deeply satisfying about that mechanical, hands-on part of the work. Because that's how I got to learn the skills.

Today, AI can do much of it.

Whether you're new to the field or already an experienced data professional, it might feel a little unsettling. You might be wondering whether your skills still matter.

I think about it too.

So here's my honest take. No hype, no doom.

The short answer: it depends. (I know, I know. Bear with me. 😉)

---

First, let's talk about what AI actually looks like inside *real* companies

Most of the AI demos you see online are built on toy projects. 5 clean-ish tables. Perfect documentation. Good structured metadata.

In that environment, yes, AI can be incredibly powerful.

But real company data systems are a completely different story.

In one project, I spent 1,5 years at a bank building a data pipeline that connected 5 source systems with hundreds of tables and millions of data records. Legacy databases, complex file structures, complex dependencies, with years of business logic stacked on top.

In another project, our team spent weeks, sometimes months talking with the doctors at the hospital we worked for, just to get the right data in place.

The real question isn't "can AI write Python code?"

It's "can AI understand the full context of your data sources, your business processes, and your internal systems, and build something production-ready from that?"

Right now, well, not yet.

And beyond the technical complexity, there's another thing people forget: technology moves fast, but companies move slowly.

You'll still find teams running critical analyses in Excel. Legacy data warehouses that have been running for 15 years. Huge chunks of company data that haven't been integrated anywhere. Becoming an "AI-driven company" overnight just doesn't happen. 🤯

It takes years of migrations, new pipelines, governance, training, and teams figuring out how everything fits together.

When someone says "AI will replace tech jobs in 2026", I genuinely think they've either never worked inside a real company, or they're trying to sell you something.

---

So what actually happens to data teams?

Here's how I think about it. AI tools are making data teams faster and more efficient. That gives companies a choice:

1. Reduce the team size: same output, fewer people
​
​2. Keep the same team, but now they can do far more with it, including unstructured data like customer reviews, PDF documents, images, and call transcripts that companies have been sitting on for years but never really knew what to do with. Until LLMs came along, it was basically untouchable for most data teams.
​
​3. Expand the team: because clean, well-prepared data is now critical not just for analytics, but for AI itself. If you feed garbage data into an AI system, you get garbage results. Someone still needs to build and maintain the foundations.

My honest read? Most companies won't go with option 1. Data teams are small, high-impact, and deeply embedded in how the business runs. I think the most risk is that companies may temporarily freeze hiring juniors.

Options 2 and 3 are where I'm seeing things move.

---

Now let's get specific: which tasks are actually at risk?

Rather than guessing at the job level, it's more useful to look at individual tasks. For each task, if you ask: can AI do this today? And can it in 5 - 10 years?

Here's my take. I summarized in tables below:

For data analysts:

Answering ad-hoc questions is probably the first to go fully. It is the most repetitive part of the job and AI can already handle it if business users learn to prompt.

The analysts most at risk are the ones whose entire day is SQL queries and simple dashboards with no stakeholder interaction. That work will be automated.

The good news? 87% of analysts say they feel more strategically valuable than ever (Alteryx, 2025). The role is not dying, it is shifting upward.

For data engineers:

Pipeline monitoring and ops is first to go.

Detecting failures and triggering retries is well-structured work AI already handles in parts today. Data modelling and architecture are among the safest tasks in any data role - they require too much business context and long-term judgment for AI to own.

The surprise? Preparing data for AI and LLMs is a genuine growth area. As companies adopt AI, demand for engineers who can feed it clean, well-structured data is actually going up, not down.​

For data scientists:

​Data scientists spend 45% less time on data cleaning when using AI-augmented tools. So if that is most of your job, it is worth paying attention.

Framing the right business problem is the hardest thing to automate and also the most valuable thing a data scientist does. It requires domain knowledge, stakeholder trust, and strategic thinking AI simply does not have.

And if you want to future-proof your career, building and fine-tuning LLM systems is the fastest-growing skill in the field right now.

---

The bottom line

If your entire value is writing basic SQL, building simple dashboards, and answering the same questions on repeat, then yes, that work is at risk.

But if you understand the business, have the domain knowledge, communicate with stakeholders, know how to model data, and how it impact business - you're going to be very hard to replace.

The honest truth is: you're worried about AI, you're probably doing more than you give yourself credit for.

Have a great week ahead! 🙌
Thu

---

P.S: Work with me:

If you want a comprehensive course from Python fundamentals to building AI applications, check out my Python for AI Projects course.

You'll join a community of 450+ learners who are building their projects while getting direct access to me and supporting each other along the way.

​🔗 Learn More​

---