Last Tuesday morning, I was stuck.
I had a spreadsheet of mutual fund performance data — 47 rows, incomplete NAV entries, scattered across three different CSV files from three different fund houses. My task was to build a comparative analysis for a client pitch at Morningstar. Normally, this takes me about 4 hours: manual data cleaning, pivot tables, cross-referencing, the whole tedious dance.
Instead, I opened ChatGPT, pasted a messy sample of my data, and asked: "Can you write me Python code to clean this and create a consolidated sheet with year-on-year returns?"
Fifteen minutes later, I had working code. I ran it. It worked. My 4-hour task became 45 minutes.
I should have felt triumphant. Instead, I felt something closer to unsettled.
That moment — sitting in my cabin at the Morningstar office in Bandra, 8:45 AM, staring at AI-generated code that actually solved a real problem — marked the beginning of how I think about work now. Not "this is the future," the way tech articles scream it. But rather: "things have actually shifted, and I need to figure out what that means for someone like me."
I've spent the last eight months intentionally using ChatGPT, Claude, and other AI tools in my daily work as a data analyst. Not because it's trendy. But because ignoring it felt professionally reckless. I wanted to understand what's real, what's hype, and what it actually means for Indian knowledge workers like us who are already juggling inflation, career anxiety, and the pressure to stay relevant.
Here's what I've learned.
1. The Obvious Productivity Gains Are Real (But They're Also Dangerous)
Let's start with what everyone talks about: speed.
ChatGPT makes you faster. Demonstrably. I can now:
- Write Python scripts for data cleaning in 1/3 the time
- Draft email explanations to clients that would have taken me 20 minutes in 3 minutes
- Generate skeleton SQL queries for database extractions
- Create first-draft presentations with talking points instead of starting from blank slides
These aren't trivial gains. Over a month, we're talking about 12–15 hours of saved time. That's a full working day every four weeks. For someone commuting from Kalyan to Mumbai every morning, that's real freedom.
But here's the dangerous part: speed without thinking is just expensive mistakes happening faster.
The Confidence Problem
In my second week using ChatGPT, I ran a Python script it generated without actually understanding what the code was doing. It produced a number. I sent that number to my manager. It was wrong — not slightly, but fundamentally. The code had made an assumption about data structure that wasn't true for my dataset.
I hadn't checked. I'd just trusted the speed.
Now I verify everything. I run the code on test data first. I ask the AI to explain what it's doing. This adds time back — sometimes 40% of the time I initially saved. But it's time spent correctly, not just quickly.
The trap is that ChatGPT is confident even when it's wrong. It doesn't hedge. It doesn't say, "I'm less sure about this part." It just generates plausible-sounding output. And when you're tired (which, if you're commuting 2+ hours daily like I do, you often are), you accept the plausible output because it's there and it works and you have three other deadlines.
When Speed Matters, and When It Doesn't
I've started categorizing my work into two buckets:
Speed-matters tasks: Formatting reports, writing routine email templates, cleaning obvious data errors, creating presentation outlines, summarizing research papers. Here, AI is genuinely transformative. I get the deliverable done faster, and the quality is high enough that I spend maybe 10 minutes refining it instead of 90 minutes creating it from scratch.
Quality-matters tasks: Building new analytical models, making recommendations on fund allocations, writing original research, anything that requires me to defend my thinking. Here, ChatGPT is a research assistant and a thought partner, not a solution generator. It helps me think through problems faster, but the core work is still mine.
If you blur these categories, you'll produce work that looks fine for about three weeks until someone notices the foundation is hollow.
2. The Real Win Is Delegating Boring, Repetitive Thinking
This is less sexy than "ChatGPT will replace your job," but it's more true for people in our position.
Before ChatGPT, when I needed to analyze market data, the workflow looked like this:
- Export raw data (boring, necessary)
- Clean and validate it (boring, necessary)
- Explore patterns and generate charts (partially interesting)
- Write up findings and recommendations (actually interesting)
Steps 1–2 took 40% of my time. Steps 3–4 took 60%, but they mattered more — that's where judgment and insight lived.
Now, with AI handling steps 1–2 automatically, I'm spending maybe 70% of my time on steps 3–4. The work that actually requires me.
And honestly? That's changed how much I enjoy my job.
I used to come home from the Morningstar office exhausted from data wrangling. Now I come home exhausted from thinking hard about what the data means. Those are different types of tired. One feels like progress; the other feels like bureaucracy.
The Compound Effect on Career
Here's what I didn't expect: when you offload the repetitive parts, you automatically end up doing more interesting work. Because suddenly you have time for it.
I've started writing deeper analysis on fund performance trends — work I would have deprioritized two years ago because I was too buried in data cleaning. My manager noticed. I got assigned to a special project analyzing market regime changes. That's the kind of work that gets you promoted, or at least makes your resume actually interesting.
Meanwhile, the person who's still doing data cleaning manually? They're stuck in that loop. Not because they're less smart, but because they're spending 70% of their time on steps 1–2.
This is why I don't actually think ChatGPT will replace most knowledge workers in India in the next 3–5 years. But it will absolutely stratify us. The people who learn to use it effectively will move into higher-judgment roles. The people who ignore it will get stuck doing commoditized work.
3. New Skills Are Emerging (And They're Not What You'd Expect)
When people talk about "upskilling for the AI era," they usually mean learning to code, or learning data science, or getting certified in machine learning.
That's partially true. But the more useful skill I've learned is something much simpler: asking better questions to AI.
Prompt engineering isn't fancy. It's just: be specific, provide context, ask for what you actually need, check if the output makes sense.
But almost nobody does this naturally.
Most people ask ChatGPT vague questions and get vague answers and then blame the AI. When actually, the AI just needed better input.
Example from my work:
Bad prompt: "Write Python code to analyze mutual fund data"
Good prompt: "I have a CSV with columns: Fund Name, NAV, Date, AUM. I need to calculate the 1-year, 3-year, and 5-year CAGR for each fund, excluding funds with less than 1 year of history. The dates are in DD/MM/YYYY format. Write Python code to do this and export results to a new CSV with columns: Fund Name, 1Y CAGR, 3Y CAGR, 5Y CAGR."
The second one takes 10 seconds longer to write but gets me a usable solution. The first one wastes 20 minutes of back-and-forth.
The Meta-Skill: Knowing When Not to Use It
Actually, the most underrated skill is knowing when ChatGPT is the wrong tool.
I don't use it for:
- Strategic decisions — it'll generate plausible-sounding advice that reflects its training data, not your specific situation
- Original research — it'll confidently synthesize existing information but won't generate new insights
- Sensitive client communication — using a draft is fine; sending an AI-generated email wholesale to a client is lazy
- Work that's supposed to be hard — if the point of the task is to develop your thinking, AI skips the development
People who use ChatGPT for everything look busy but produce mediocre work. People who use it strategically produce better work faster.
4. The Career Anxiety Is Justified, But Not in the Way You Think
I'm not worried that ChatGPT will replace me as a data analyst.
I'm worried that it'll create a glut of people who think they can do data analysis. Someone with zero statistics knowledge can now write Python code that produces numbers that look professional. Those numbers might be garbage, but they'll look fine to a non-technical manager.
So there'll be more competition, but for a lower tier of work.
The real anxiety — the one I feel more acutely after eight months of using these tools — is this: What happens when everyone has access to the same AI?
If I'm using ChatGPT to write analysis, and so is my competitor across India, and so is someone in the Philippines, the differentiator stops being "can you write good analysis" and becomes "can you think of analysis that nobody else has thought of yet."
That's harder. It requires original judgment, reading patterns in data that others missed, making a bet on what comes next.
It's also what I actually got into this field to do.
The Economic Reality for Indian Workers
Here's something I think about a lot while sitting on the 6:47 AM Kalyan train to Mumbai:
ChatGPT is a tool that amplifies judgment. If you have good judgment, you become more productive and more valuable. If you don't, you become faster at producing mediocre work, which actually makes you less valuable than before.
For Indians in knowledge work, especially in the post-pandemic era when we're already competing globally, this is significant. Our competitive advantage was never "we're cheaper" — that advantage keeps eroding as AI commoditizes work. Our advantage is "we can think our way out of hard problems."
ChatGPT is either going to accelerate that advantage or destroy it, depending on how we use it.
| Work Type | ChatGPT Impact | Your Strategic Move |
|---|---|---|
| Routine analysis, reports, templates | 5-10x faster | Use it, redirect freed time to higher-value work |
| First-draft documentation, emails | 3-5x faster | Use it as a starting point, always refine for your voice |
| Original research, insights, recommendations | Modest (1.2-1.5x) | Use it for research and structure, but the thinking is yours |
| Strategic decisions, client relationship management | Negative (risky) | Avoid it, develop your own judgment here |
5. What Actually Matters Now
After eight months of integration, here's what I genuinely believe matters for someone in my position:
Judgment Beats Speed
I can generate output faster now. But I can also be confidently wrong faster. The only thing that saves me is judgment — knowing which outputs to trust, which to question, and which to throw away entirely.
That judgment comes from experience and thinking deeply about your field. ChatGPT can't give you that. It can only help you deploy it faster.
Irreplaceable Work Is About People, Not Data
The work I do that nobody could replicate with ChatGPT isn't my data analysis. It's the relationships I've built with clients, the patterns I've noticed because I've been watching the market for eight years, the recommendations I make because I understand the person asking for them.
AI excels at taking what's known and recombining it. It's terrible at understanding what someone actually needs when they haven't explicitly said it. That's still a human skill.
Continuous Learning Isn't Optional Anymore
I used to feel like I could coast for a few years on what I learned in my Economics degree and early career. That window is closed.
Not because I need to learn machine learning or whatever the trend is. But because the tools are changing how work gets done, and if you're not actively thinking about that, you'll wake up in three years unable to understand what your junior colleagues are doing.
My Perspective
I've been thinking about whether to buy a new laptop — my current one is four years old and getting slow. After I started using ChatGPT, my budgeting priority shifted.
I used to think: save ₹50,000 for a laptop, use it for 4 years, write it off against wear and tear. Clean financial decision.
Now I'm thinking: invest in a good laptop now because my productivity gain from AI tools is only realized if I have hardware that can run them properly. A slow machine negates the speed benefit.
So I'm saving ₹1,50,000 instead of ₹50,000 to buy something really good — not premium, but genuinely capable. I'm also setting aside ₹5,000 a year for ChatGPT Plus and other tools.
What surprised me: this financial decision wouldn't have made sense six months ago. But now it does, because the math has shifted. The bottleneck used to be my thinking time. Now it's my computing power and access to good tools.
I got that part wrong initially. I was so focused on learning new skills that I missed the simpler point: invest in the infrastructure that lets you use these tools properly.
Final Thoughts
ChatGPT isn't magic. It's a tool that's genuinely useful if you understand what it's good for, and genuinely dangerous if you don't.
For people like us — Indian millennials in knowledge work, trying to build careers, trying to figure out how to stay relevant — the answer isn't to panic and also isn't to ignore it.
It's to be pragmatic. Use it where it saves you real time. Learn to ask it good questions. Keep thinking for yourself on the work that matters. And remember that the bottleneck in most knowledge work isn't anymore how fast you can produce output — it's whether that output is actually right and actually useful.
That last part? That's still on you.
The train from Kalyan to Mumbai is still full of people on their phones reading news. Most of them aren't thinking about how their work is changing. But the ones who are — the ones who are sitting there asking "how do I use this properly?" — those are the ones who'll be fine.
Maybe better than fine.
Dattatray Dagale
Data Analyst • Blogger • Mumbai
I'm a data analyst from Kalyan, Maharashtra, working at Morningstar. I write about personal finance, career growth, and everyday life for Indian millennials — the stuff I wish someone had told me earlier.
Written by Dattatray Dagale • 01 August 2026
.gif)
.png)
.png)
0 Comments