Most writing about AI and analysts is either a threat narrative or a tool listicle. Neither is much use if you are an analyst deciding what to do differently on Monday.
Here is the practical version: which parts of your day change, which skills go up in value, and the specific trap worth avoiding.
What genuinely gets faster
- Writing the query. Describing what you want in plain English and getting working SQL is a real time saving, particularly against unfamiliar schemas. Read it before you run it — a subtly wrong join produces a plausible number, not an error.
- Cleaning and reshaping. Format normalisation, deduplication, pivots, joins across messy sources. Ask for the transformation logic, not just the output.
- First-pass exploration. "What is unusual here" now returns something useful immediately, which is genuinely a better starting point than a blank notebook.
- Chart production. From description to a formatted chart in seconds.
- Writing the commentary. Turning a result into a paragraph a manager will read — a task many strong analysts find disproportionately slow.
Taken together that is a substantial share of a typical analyst week.
The trap
Here is where it goes wrong, and it is worth being blunt about.
If AI makes producing analysis three times faster and you respond by producing three times as much analysis, you have not become more valuable. You have increased the volume of output competing for the same finite attention from decision-makers — and average quality falls, because the scarcity that used to force prioritisation is gone.
The analysts whose value is rising are doing the opposite: producing less, on questions that matter more, with more rigour per piece. The reclaimed hours go into understanding the business, pressure-testing conclusions, and following up on whether anyone acted.
That is a genuinely uncomfortable reframe, because "produced more reports" is easier to put in a performance review than "prevented a bad decision."
Which skills appreciate
Knowing which question to ask
The stakeholder asks for last quarter's numbers by region. The valuable analyst works out they are trying to decide where to open next, and that the useful analysis is different from the one requested. AI cannot do this — it answers what you asked. See the methods map.
Judging whether a result is trustworthy
This has become the core skill. AI-generated analysis fails differently from human analysis: well-formatted, articulate, internally consistent and wrong. Catching that requires knowing the data well enough to notice when a number is impossible — not just implausible.
Causal reasoning
Distinguishing "these move together" from "this caused that" is where analysts add the most defensible value, and where AI most convincingly overstates. See correlation vs causation.
Communicating to a decision
AI writes competent summaries. It does not know that your CFO cares about cash timing rather than margin, or that the ops director will dismiss anything without a named owner. That context is yours.
Which skills depreciate
Worth naming honestly rather than pretending everything is additive.
- Memorising syntax. Knowing SQL window-function syntax by heart is worth much less than it was. Reading and verifying it is worth more.
- Chart-building as craft. Producing a clean chart is now trivial. Choosing the right chart for the question, and titling it with the finding rather than the variable, is not.
- Being the person who can pull the data. If your role was largely a self-service bottleneck, that moat is eroding fast. Move toward interpretation.
A practical way to use AI without losing rigour
- Never accept an answer without the method. Ask for the query, the transformation and the assumptions. If you cannot follow it, you cannot defend it.
- Ask what would change the conclusion. "What alternative explanation would make this wrong?" is a five-minute habit that catches real errors.
- Sanity-check against something you already know. Run it on a period where you know the answer.
- Keep a reference set. A handful of metrics you know cold, to check any new pipeline against.
- Write the decision, not the analysis. Lead with what should happen; put the method below for anyone who wants it.
Where the role is heading
The direction is fairly clear: away from producing analysis and toward being accountable for whether decisions improved. That is a bigger job, not a smaller one, and it needs more business context rather than less.
Analysts who move that way are becoming more valuable, not less. Analysts who stay in the produce-charts-on-request lane are competing with a tool that does it instantly and does not take leave. The broader shift in data science points the same direction.
Our AI Analytics programme is built around this — methods, judgement and communication with AI tooling, for working analysts and business teams. HRD Corp SBL-KHAS claimable for eligible Malaysian employers.
