On 24 February 2026, Nasscom put out its Annual Strategic Review. The industry total came to roughly $315 billion for FY26, up 6.1% on the year (YourStory, Feb 2026). Headcount was a different story. The sector added about 135,000 people, taking the total to nearly 5.95 million, and that works out to about 2.3% growth (The Core, Feb 2026). It was also the first year Nasscom estimated AI revenue on its own, at $10–12 billion (Business Standard, Feb 2026).
Revenue grew at roughly two and a half times the rate of hiring. For anyone graduating in 2026 or 2027, that gap matters more than the headline number. The industry is still hiring. It just needs fewer people per dollar of work, and the coverage around the review (The Core, Feb 2026) put campus hiring among the areas that have slowed.
What the gap actually means
Most of the industry's revenue still comes from familiar services work: application maintenance, testing, support, data engineering, migrations. A growing share of that work is now done by a smaller team with AI tools in the loop. Somebody drafts the test cases with a model and somebody reviews them. A support ticket gets summarised automatically, and an engineer decides whether the summary is right.
So the entry-level role is being redesigned, not removed. Where a fresher used to spend the first year writing boilerplate and running manual test passes, the same seat now involves three things:
- Framing the task so a model produces something usable: clear inputs, examples, constraints and an output format someone else can check.
- Checking the output against data, specs or a test suite, and catching the confident mistakes.
- Wiring small pieces together: a Python script that pulls a file, cleans it, calls a model or an API, and writes a report someone actually reads.
None of this is exotic. It is ordinary software and analysis work with a new tool in the middle. The people who get hired are the ones who can show they did it on a real problem, not the ones with the longest list of course certificates.
A realistic example
Take a second-year analyst on a finance operations team at a services company in Pune. Every Monday they receive 40 vendor invoices as PDFs and a spreadsheet from procurement. The old job was copying fields by hand and flagging mismatches.
The newer version of the same job looks like this. A short Python script reads the spreadsheet with pandas. A model extracts invoice number, GSTIN, amount and due date from each PDF into a fixed JSON format. The script compares the two, writes every mismatch to a sheet, and marks the five where the model was least certain for a human to open. The analyst's real work is the last mile: deciding which mismatches are genuine, fixing the prompt when the model keeps misreading one vendor's layout, and explaining the numbers to the team lead.
That is two or three days of learning for someone who already knows basic Python, and it is exactly the kind of thing you can put in a portfolio.
Skills that are rising
Based on where the work is moving, these are worth your time in roughly this order:
- Python and pandas for everyday data work. Reading files, cleaning messy columns, joining tables, handling errors without the script falling over.
- Prompting as specification. Writing instructions with examples and a strict output format, then testing them on 20 cases instead of one.
- Evaluation. Deciding what "correct" means for an AI output and measuring it, even crudely, with a labelled sample.
- Retrieval and tool use. Once you are comfortable, learn how models answer from your own documents (RAG) and call functions or APIs (agents). This is where the AI revenue line in that Nasscom number comes from.
- Judgement about risk. Knowing what data should never go into a public tool, and when an AI answer needs a human sign-off.
How to start
Week 1–2. Install Python, pick one dataset you care about (cricket scores, your college results, UPI transaction exports from your own bank app with personal details removed), and answer five questions about it with pandas. Write the answers in a short README.
Week 3. Take one boring task you or a friend does every week and write down each step. Try to automate two of the steps with a script and a model. Keep the version that fails; it teaches you more.
Week 4 onwards. Build an evaluation for that workflow: 20 inputs, the answer you expected, and what the system produced. Fix the worst failure, run it again, and note what changed.
Here is how our courses map to that path.
Python for Data Analysis (free) covers the base layer. You work through data types and control flow, functions and modules, and the lesson on file I/O and error handling, which is the part most self-taught scripts skip. Week 2 goes deep on NumPy arrays and pandas, including time series. Week 4 includes a lesson on automation and reproducibility, and the course ends with a data analysis capstone you can show.
AI for Professionals is the workflow course. Week 2 covers prompting for business use cases and has you build a prompt portfolio. Week 3 has a lesson on building AI workflows without code and an assignment to build a working AI workflow. Week 4 covers AI governance and ethics and the business case for AI, which helps when you need to explain your work to a manager.
Generative AI Masterclass is for later, once the first two feel easy. It covers RAG architecture and RAG evaluation in production, building agents with tool calling, and a week on safety, evaluation and LLMOps. The assignments include a document Q&A system and a multi-tool research agent.
Who this is for: students and early-career people in IT services, GCCs, analytics or operations who can already write a little code, or are willing to start, and want to do the work that is replacing routine tasks.
Who this is not for: anyone looking for a shortcut certificate to put on a CV without building anything. Also not the right fit if you want research-level machine learning; these courses are about applying models, not training them from scratch.
The honest takeaway
Nasscom's numbers say the industry is growing and still adding people, just more slowly than its revenue. That leaves less room for roles that only follow instructions and more for people who can set up a workflow, check it and explain it. You can start building that skill this month with a laptop and a free course.
Start with Python for Data Analysis, then move to AI for Professionals when you are ready to build workflows. If you want live webinars, session recordings and the members' WhatsApp community while you learn, Getskilled membership is ₹99 for 90 days.
