How AI is used in data science: what actually happens on the job
Updated on July 18, 2026 6 minutes read
A data scientist at a Wellington power retailer wants to know why customers churn. They pull two years of billing data, plot it, and spot a pattern. Then they hand the messy part — predicting who leaves next month — to a machine learning model. That single project is how AI is used in data science in one sentence: the human does the thinking and the framing, the model does the pattern-matching at a scale no person could manage by hand.
If you've been trying to work out whether data science and AI are the same thing, rivals, or teammates, the honest answer is that they overlap heavily and it's rarely a case of one replacing the other. Let's pull the two apart so the picture is clear.
What data science and AI each actually mean
Data science is the whole process of turning raw data into decisions. That means collecting it, cleaning it (which eats far more time than anyone expects), exploring it, building models, and explaining the results to people who don't care about the maths. It's part statistics, part programming, part being able to tell a straight story with a chart.
Artificial intelligence is narrower. It's the set of techniques that let a computer make predictions or decisions without being told the exact rules — things like classification, forecasting, and language models. Machine learning is the engine most people mean when they say "AI" at work.
So AI sits inside data science as one very powerful tool in the kit. A data scientist might spend a morning writing SQL and a query for a dashboard, then spend the afternoon training a model. Same person, same project, two different skills.
Where AI shows up in a data science project
Here's the churn example again, but slowed down so you can see where AI does the heavy lifting.
The data scientist starts with a spreadsheet-style table: one row per customer, columns for how long they've been with the company, how many times they rang support, whether their last bill jumped. None of that is AI yet — it's just careful data prep and asking good questions.
The AI part kicks in when they feed those columns into a model and ask it to learn the relationship between customer behaviour and cancelling. The model finds patterns a human would miss, like a specific combination of a price rise plus two support calls being a strong warning sign. Once trained, it scores every current customer with a churn probability, and the retention team rings the high-risk ones first.
AI also quietly speeds up the boring parts. Language models can now draft the SQL, suggest which chart fits the data, and summarise a long results table in plain English. Used well, that frees a data scientist to spend more time on judgement calls — which is exactly the part a model can't do for you. If this sounds like work you'd enjoy, our data science and AI bootcamp curriculum walks through this full workflow with hands-on projects.
AI vs data science: which is "better"?
This question comes up constantly, and it's a bit like asking whether an engine is better than a car. One is a component of the other. Still, if you're deciding where to point your learning, the comparison below helps.
| Data science | Artificial intelligence (as a job focus) | |
|---|---|---|
| Core question | "What is the data telling us, and what should we do?" | "How do we get a machine to predict or decide accurately?" |
| Daily work | Cleaning data, analysis, visualisation, stakeholder reporting | Building, training, tuning and deploying models |
| Key tools | Python, SQL, pandas, Power BI, Tableau | Python, scikit-learn, TensorFlow or PyTorch |
| Strongest for | Broad roles that touch the whole data lifecycle | Deep, specialised model-building roles |
| Maths load | Moderate — stats and probability | Heavier — linear algebra, optimisation |
Neither is "better" in the abstract. If you like variety and talking to people, data science suits you. If you enjoy going deep on algorithms, an AI or machine learning slant fits better. In practice most NZ job ads blur the line anyway, and having both makes you more hireable across Auckland, Christchurch and Wellington.
Does a data scientist actually work with AI?
Yes, and increasingly it's expected. Look at data scientist and analytics roles advertised by NZ banks, telcos, government agencies and firms like Xero, and you'll see machine learning listed alongside SQL and statistics. A data scientist doesn't need to invent new algorithms — that's more the domain of a research scientist or a specialised ML engineer — but they're expected to use existing models competently and know when a model is fooling itself.
There's a useful distinction here between the analyst end and the engineering end. A data scientist trains a model to answer a business question. A machine learning engineer then makes that model run reliably in production, at scale, without falling over. The two roles hand work back and forth. If you're weighing up the roles, our plain-English breakdown of the full range of tech courses at Code Labs Academy shows how each specialism connects.
Can AI replace data science?
Short version: no, though it will keep changing the job. AI tools are brilliant at the mechanical parts — writing boilerplate code, spotting patterns, generating first-draft charts. They're poor at the parts that make data science valuable: deciding which question is worth asking, noticing when the data is biased or incomplete, and explaining to a non-technical manager why a "promising" result is actually noise.
A model will happily give you a confident answer to a badly framed question. Catching that is human work. Think of AI as a very fast junior who never gets tired but has no common sense — someone still has to check the output and take responsibility for the decision.
What's more likely than replacement is that data scientists who use AI well will outpace those who don't. The floor for repetitive tasks is rising, so the skill that pays is judgement plus the ability to steer these tools.
How to start in New Zealand
You don't need a maths PhD to begin. A practical path is to get comfortable with Python and SQL, learn enough statistics to avoid fooling yourself, then build two or three small end-to-end projects using real datasets — Stats NZ open data is a good local source. Employers here care far more about a portfolio they can look at than about a long list of course names.
Structured learning speeds this up, especially if you want feedback on your projects and a clear sequence rather than a pile of random tutorials. If you're juggling a job, the self-paced data science and AI track lets you work through it around your own schedule.
The takeaway is simple: data science and AI aren't competitors — AI is the sharpest tool in the data scientist's kit, and the people who learn to wield it will be the ones NZ employers chase. If you're ready to build that skill set properly, compare your options and see the data science and AI course details and pricing to find a format that fits your life.
