Do You Still Need A Data Scientist When AI Writes The Code?

Every few weeks a business owner asks us why they should still pay a data scientist when AI can write code, build charts, and answer questions in plain English. Sometimes they should. Often they should not. The work is in telling the two situations apart, and this post gives you the test we use.

What AI Actually Changed

AI did not remove the need for thinking about data. It removed a lot of the typing. Tasks that used to eat a junior analyst's week, cleaning a messy file, writing a first draft of a query, sketching a chart, now take minutes. That is real, and it is a gift to small businesses that could never afford a full data team.

But AI is confident whether it is right or wrong. It will happily build you a handsome forecast on a number that means the opposite of what you think it means. It does not know that your December figures are inflated by a one off government contract, or that two branches record a sale on different days, or that the field labelled revenue is actually revenue before returns. Someone has to know the business and the data well enough to catch that. That someone is doing data science, whether they hold the title or not.

Think of AI as an eager graduate who types at a thousand words a minute, never pushes back, and never says a job is beyond them. That is enormously useful when you know exactly what you want and can check the result. It is dangerous when you cannot, because the mistakes arrive polished, formatted, and sounding certain. The skill that matters now is less about producing the analysis and more about knowing whether to trust it.

When You Do Not Need To Hire One

If your questions are mostly descriptive, what sold, where, and to whom, and your systems are clean, you may not need a full time data scientist at all. Modern tools and a capable analyst can carry you a long way. Hiring a senior specialist to build dashboards is like hiring an architect to hang a door.

• You need reporting and dashboards, not prediction.
• Your data lives in one or two clean systems.
• The questions repeat and rarely change.

Say a single-location pharmacy wants to know which products sell best by day of the week and which suppliers are slipping on delivery. That is a reporting job. A sharp analyst with a good tool, or an owner who is comfortable in a spreadsheet, can answer it and keep answering it every month. Paying a senior data scientist a senior salary to run that is money spent on horsepower the job never uses. Match the hire to the question, not to the fashion.

When You Genuinely Do

A data scientist earns the salary when the question is hard and being wrong is expensive: predicting which customers will leave, pricing a product nobody has sold before, deciding which of forty branches to open a forty-first beside. These are judgement problems dressed up as maths problems, and the judgement is the part AI cannot supply.

The tell is stakes plus uncertainty. A lender deciding who to approve is not asking what happened last month. It is betting money on what a stranger will do over the next two years, and a model that is quietly wrong loses real money on every loan it touches. One lender we worked with rebuilt that credit model properly and cut defaults by 18 percent, the kind of result that pays a specialist's salary many times over and is nearly impossible to reach by prompting a tool that has never seen your loan book. When the decision is expensive and the answer is genuinely unknown, you are buying judgement, and judgement is what you should insist on.

• The decision carries real money or real risk if it goes wrong.
• The answer depends on predicting behaviour, not describing the past.
• The data is messy, scattered, or easy to misread without context.

The Prestige Hire Trap

A lot of bad data hires start with ego rather than need. A competitor announces a head of data science, so the board decides it needs one too, and a big salary gets approved before anyone has written down a single question worth that salary. The new hire arrives, finds the systems are a mess and the questions are all reporting, and spends year one doing plumbing and dashboards that a good analyst and a tool could have handled for a fraction of the cost. Nobody is happy, and the resume value of the title quietly becomes the main thing the role produces.

Test the need before you post the job. Take your single hardest, most expensive decision and run it as a small paid pilot, with a hire on contract or a partner, before you commit to a permanent salary. If the pilot moves a real number, you have proof the role pays and a template for what it should do. If it does not, you have saved yourself a year of payroll and learned that your bottleneck was somewhere else. Either way you are deciding on evidence, which is the whole point of hiring a data scientist in the first place.

The Option Most People Miss

For most Caribbean businesses the answer is not full time hire or nothing. A senior data scientist commands a salary few firms can justify for work that comes in waves, and the hiring itself is a gamble. The talent pool is thin, the good ones are expensive, and a wrong hire can sit on payroll for a year building dashboards nobody asked for. Even when you find the right person, a single hire is a single point of failure the day they take another job.

A partner on retainer gives you the same capability when you need it, and costs nothing when you do not. You get the senior judgement without the year round payroll and the hiring risk, and you get a team behind that judgement rather than one person's blind spots. When the churn model is built and running, you are not paying a specialist to fill the quiet months inventing work.

There is a fourth option worth naming: train your own people. An analyst who already knows your business, taught to use the new tools well and to spot when a result cannot be trusted, is often the best value of all. They stay, they carry the context, and they get sharper every year. The tools are cheap now. The judgement to drive them is what you are actually building.

So run the test. Data science matters more in the AI age, not less, because everyone now holds the tools and few have the judgement to use them well. What changes is how you buy that judgement: hire it on staff, rent it on retainer, or grow it in the people you have. Pick by two numbers you already know. How often do you face a decision this hard, and what does a wrong call cost you when you get it? Rare and cheap points to a tool and an analyst. Frequent and expensive justifies a hire. Somewhere in between, which is where most businesses live, points to a partner.

How To Decide In The Next Week