How to Break Into Remote Data Jobs – When AI Has Changed What “Data Skills” Means
Image by the digital artist
Two years ago, if you wanted to break into a remote data job, the advice was simple. Learn SQL. Learn Python. Take a course on statistics. Build a Tableau dashboard. Apply to junior data analyst jobs and hope someone gives you a shot.
That advice isn’t wrong. It’s just not enough anymore.
AI hasn’t eliminated data jobs. If anything, companies need more people who can work with data than ever before. But it has compressed what counts as entry-level skill, expanded what hiring managers expect, and created entirely new categories of data work that didn’t exist when most career guides were written.
If you’re trying to break into remote data jobs right now, here’s what’s actually changed and what to do about it.
The baseline has shifted upward
The tasks that used to define a junior data analyst like pulling queries, cleaning spreadsheets, building basic dashboards, can now be done faster by someone with AI assistance and no formal data training. A product manager who knows how to prompt Claude or ChatGPT can write a decent SQL query without ever having taken a course. A marketing lead can clean and visualise a CSV in minutes using Code Interpreter.
This doesn’t mean SQL and Python are worthless. They’re still essential. But they’re no longer differentiators. They’re prerequisites. Listing “SQL, Python, Tableau” on your CV puts you in the same bucket as everyone else. It doesn’t tell a hiring manager what you can actually do with those tools that AI-assisted non-specialists can’t.
The candidates getting hired into remote data roles right now are the ones who can demonstrate something beyond tool proficiency: the ability to take a messy, ambiguous business question and turn it into a clear answer that drives a decision. That’s the part AI can’t do on its own, and it’s the part most portfolios completely miss.
The new data roles worth targeting
The traditional “data analyst” title still exists, but the landscape has split into more specific roles, several of which are newer, less crowded, and more AI-friendly than the classic analyst path.
Analytics Engineer. Sits between data engineering and data analysis. You build and maintain the data models that analysts and business teams rely on. Tools: dbt, SQL, a cloud warehouse like BigQuery or Snowflake. This role has exploded in demand over the past two years because companies realised their data was a mess and someone needed to fix the plumbing. It’s technical but doesn’t require software engineering experience, and it’s almost always remote.
Product Analyst. Focused specifically on how users interact with a product. You’re answering questions like: where do users drop off, which features drive retention, what should we build next. Tools: SQL, Amplitude or Mixpanel, basic statistical reasoning. Product analyst roles tend to sit inside product teams rather than a central data team, which means more direct business impact and more interesting problems.
AI/ML Data Analyst. A growing category where you’re evaluating, testing, and improving AI model outputs rather than building models from scratch. Companies training or fine-tuning models need people who can assess data quality, design evaluation frameworks, and identify where the model is failing. This doesn’t require a machine learning background, but it does require sharp analytical thinking and attention to detail.
Revenue / Business Operations Analyst. Combines data analysis with operational work. You’re building the dashboards, reports, and forecasting models that leadership uses to make decisions. Tools: SQL, spreadsheets (seriously — advanced spreadsheet work is underrated), a BI tool, and increasingly some automation via Python or n8n. These roles are common at startups and scale-ups where one person needs to cover both the analysis and the systems that deliver it.
What to build in your portfolio before you apply
The portfolio advice for data jobs has always been “do a Kaggle project.” That’s fine, but Kaggle projects all look the same and they don’t demonstrate business thinking. Here’s what actually works.
Pick a real question, not a practice dataset. Find a public dataset that relates to something you genuinely care about. For example, housing prices in your city, job market trends in your industry, sports performance, public health data, whatever. Frame it as a business question: “Which neighbourhoods are underpriced relative to their amenities?” not “I ran a regression on housing data.”
Document your thinking, not just your code. The code is table stakes. What hiring managers want to see is how you approached the problem. Why did you choose this metric? What did you exclude and why? What was surprising in the data? Write a short narrative alongside the analysis — a blog post, a README, a Notion page. Show reasoning, not just output.
Use AI openly and demonstrate what it can’t do. This is the move most candidates are missing. Use Claude or ChatGPT to help with your analysis – data cleaning, initial exploration, code suggestions, and then show where you had to override it, correct it, or think beyond what it gave you. Write a section called “where AI helped and where it didn’t.” This signals to a hiring manager that you understand AI as a tool with limits, not a magic box. That’s exactly the judgment they’re hiring for.
Build one end-to-end pipeline, not five disconnected analyses. Set up a project that goes from raw data to clean model to dashboard to insight. Use a free cloud warehouse (BigQuery has a generous free tier), dbt for transformation, and a BI tool (Looker Studio is free, Metabase is open source) for the output. A single project that shows you can handle the full stack is worth more than five Jupyter notebooks that each do one thing.
The meta-skill that matters most
Every data job – data analyst, data analytics engineer, product analyst, ops analyst, ultimately exists to help someone make a better decision. The technical skills get you through the ATS. The business thinking gets you the job.
When you’re building your portfolio to get hired into remote data jobs, ask yourself for every project: could a non-technical person read this and understand what to do next? If the answer is no, the analysis isn’t finished.
AI has raised the floor on technical data skills. Anyone can write a query now. The ceiling *translating data into decisions* is still where the value lives. That’s what your application needs to prove.