AI and Data Analysts: What's Actually Happening and What to Do

The honest assessment

Data analysis sits in an interesting spot. It's the profession that arguably should have seen AI coming first, given that you lot literally work with the stuff AI runs on. And yet a lot of data analysts are still writing SQL queries by hand like it's 2019.

Here's what AI can do right now. Write SQL queries from plain English descriptions. Clean and transform datasets. Generate Python and R code for statistical analysis. Create visualisations from raw data. Identify patterns, outliers, and trends. Produce written summaries of analytical findings. Microsoft Copilot in Excel can do things that would have required a data analyst two years ago... a marketing manager can now type "show me the monthly trend for customer churn by segment" and get a chart. ChatGPT can write a complete ETL pipeline from a description of the data sources.

What AI can't do well is ask the right questions. It can answer "what's the correlation between X and Y" instantly. But it can't look at a business problem and figure out that X and Y are the wrong variables to examine, and that the real insight is hiding in Z, which nobody thought to ask about. The ability to frame analytical problems, to understand what the business actually needs to know, to challenge assumptions, and to communicate findings in a way that changes decisions... that's still human territory.

The part that's shifting fastest is the technical execution layer. Writing code, building dashboards, and running standard analyses are all being accelerated or automated. The gap between "I have a question about my data" and "I have a chart with the answer" is shrinking rapidly. This is good for businesses. It's complicated for data analysts who defined their value by their technical skills alone.

Here's the thing that should genuinely keep data analysts up at night though. AI democratises analysis. When anyone in the company can query a dataset using natural language, the data analyst's role as gatekeeper disappears. You're no longer the person with the special skills needed to access the data. You need to be the person who knows what to do with it.

Your exposure level: Medium

Medium exposure, but leaning toward the higher end of medium. The technical core of data analysis... writing queries, building reports, creating dashboards... is highly automatable. What protects the profession is the strategic layer on top.

A McKinsey analysis estimated that about 40% of data analysis tasks could be automated with current AI technology. That's not a small number. But it's the 40% that most data analysts find tedious anyway... data cleaning, routine reporting, standard visualisations. If AI handles that, what's left is the interesting work. The problem-framing. The storytelling. The "here's what this actually means for the business" conversation.

The catch is that not every data analyst does that interesting work. Some roles are primarily about producing reports that someone else interprets. If that's your role, your exposure is higher than medium. The analysts who are best positioned are those who combine technical skills with business acumen and communication ability. If you can write a SQL query AND explain to the board why the data matters AND push back when someone asks for analysis that will lead to a bad decision... you're in good shape. If you can only do the first thing, start working on the other two. Quickly.

The 90-day action plan

1. This week: let AI write your queries. Take your last five SQL queries. Describe what each one does in plain English, then ask ChatGPT or Copilot to write the SQL. Compare the output to yours. Is it as efficient? Does it handle edge cases? You'll learn where AI is good enough and where your expertise still matters.

2. Week two: automate a routine report. Pick the report you produce most frequently. Build a workflow where AI handles the data extraction, basic analysis, and first draft of the commentary. You review, refine, and add insight. Time the before and after. The time savings should be significant.

3. By day 30: learn to use AI for exploratory analysis. Instead of your usual approach, try something different. Paste a dataset description into Claude and ask "what are the most interesting questions we could ask of this data?" Use it as a brainstorming partner. i've found it suggests angles I wouldn't have considered, about 30% of the time. That 30% is worth the exercise.

4. By day 45: build your communication skills. This is the uncomfortable one. Most data analysts chose this career partly because they preferred working with data over working with people. But the ability to translate analytical findings into business decisions is the single most valuable skill you can develop right now. Take your next analysis and write it up as a one-page executive summary before building the detailed report. Practice the narrative.

5. By day 60: learn a new analytical technique. Causal inference. Bayesian methods. Time series forecasting with modern approaches. Whatever's adjacent to your current skill set but deeper. AI handles standard analyses well. It handles advanced, nuanced statistical work less well. The deeper your expertise, the more defensible your position.

6. By day 75: become the AI tools person on your team. Test and evaluate AI analytics tools. Write up what works, what doesn't, and what the team should adopt. The person who leads AI adoption in the analytics function has a very different career trajectory from the person who has it done to them.

7. By day 90: have the positioning conversation. Talk to your manager about how your role is evolving. Frame it as: "I can now produce our standard reporting in a fraction of the time. I'd like to use that capacity for deeper strategic analysis on [specific business problem]." You're volunteering for more valuable work. That's hard to say no to.

AI tools you should be using this week

• Microsoft Copilot for Work — The Excel integration is the headline for data analysts. Natural language queries against your data, automatic chart generation, formula writing, and data cleaning suggestions. If you work in a Microsoft environment, this should be your first stop.

• ChatGPT for Work — Excellent for writing and debugging code. SQL, Python, R... it handles all of them well. Also good for generating documentation, writing analysis summaries, and brainstorming analytical approaches to business problems. The Code Interpreter feature can run analysis directly.

• Claude for Work — Particularly strong at understanding complex analytical frameworks and producing nuanced written summaries. i use it when I need to think through a complex analytical problem. Describe the situation, the data you have, the question you're trying to answer, and let it suggest approaches.

• Perplexity for Research — Useful when you need context for your analysis. Industry benchmarks, market trends, or methodological best practices. Faster than trawling through academic papers and gives you cited sources.

What to say in meetings

When someone asks if data analysts are being replaced: "AI handles the data extraction and basic reporting faster than we can. That's fine. Our value isn't in writing queries. It's in knowing which queries to write and what the answers mean for the business." Simple, true, and positions you correctly.

If a stakeholder says "can't we just ask Copilot?": "You can, and for straightforward questions it'll give you a good answer. For anything where the interpretation matters... where you need to understand statistical significance, data quality issues, or business context... that's where we come in." Don't be defensive. Be specific about your value.

In team planning sessions: "I've reduced the time on routine reporting by 50% using AI tools. I'd like to propose we spend that freed-up time on [specific strategic analysis project]. Here's the business case." Convert efficiency into impact.

If the worst happens

If you're made redundant from a data analyst role, your technical skills plus business understanding make you employable in a surprisingly wide range of roles. Data engineering, business intelligence, product analytics, marketing analytics, and data science all draw on similar foundations. The market for people who can work with data is actually growing... it's just changing shape.

Adjacent moves: business analyst, data engineer, product analyst, analytics consultant, or data science (if you're willing to go deeper on the statistical and ML side). The consulting route is particularly viable. Every company knows they should be using their data better. Very few know how. An analyst who can walk in, understand the business, connect the right data sources, and deliver actionable insights... that's a consultant worth paying for.

One thing i'd push back on if you're feeling anxious. Data literacy is becoming more important, not less. As AI puts analytical tools in everyone's hands, the need for people who can ensure data quality, interpret results correctly, and prevent bad decisions based on bad analysis is growing. You're not being replaced by AI. You're being repositioned from "the person who runs the report" to "the person who makes sure the report means something." If you can make that shift, the job market is better than the headlines suggest.