Fahad Manzur
Entry-level freelance listings have dropped from about 15% of the market to below 9%, and basic writing and translation work is down 32% year on year, according to figures reported by The Express Tribune in the same week Pakistan posted a record $4.6 billion in IT exports. For a beginner in Pakistan, that means the traditional first rung (data entry, article rewriting, simple translation) is now the most crowded and lowest-paid place to start. The better entry point in 2026 is a narrow service where you operate an AI tool for a client, check its output, and take responsibility for the result.
Those two numbers sit inside a bigger story. State Bank of Pakistan data put IT and IT-enabled services receipts at $4.6 billion for FY2025-26, up about 21% from $3.814 billion, with freelancers contributing roughly $1.6 billion, a figure Finance Minister Muhammad Aurangzeb repeated at a Lahore event on 4 September. In the same year, P@SHA found entry-level hiring at major software houses down around 25%, and employment among developers aged 22 to 25 has fallen nearly 20% since 2024. Money is flowing in. The bottom rung is being pulled up behind it.
Why beginners are hit first
The work a newcomer could realistically win in 2023 (rewriting a blog post, transcribing an audio file, cleaning a spreadsheet, translating a product description) is exactly the work a client can now get from a chatbot in thirty seconds. Clients did not stop needing it. They stopped paying a person to do it from scratch.
The pattern is global. A July 2026 Gartner survey of 110 heads of HR found that 22% of CHROs report at least one business leader in their organisation has stopped hiring for entry-level roles because of AI automation. Gartner’s explanation is useful for a freelancer: AI at most organisations is being pointed at “the less complex tasks traditionally performed by entry-level workers”, which leaves a mismatch between what a beginner can offer and what is left to buy.
Syed Azfar Hussain of the National Incubation Center Karachi told the Tribune that demand is rising for a different workforce, naming AI, cloud, cybersecurity and data engineering. That is accurate but not very helpful if you have three months, a laptop and no degree. The practical question is what a beginner can sell now that a chatbot cannot, and the answer is narrower than “learn AI”.
The same clients are still buying the same categories
One detail in the platform data matters more than the headline. Upwork’s In-Demand Skills 2026 report, built on completed-job earnings across 2025, still lists Data Entry as the second most in-demand skill in Customer Service and Admin Support, behind General Virtual Assistance, with Manual Transcription at number ten. General research, market research and executive virtual assistance are all in that top ten too.
So admin work has not vanished. What changed is the price and the shape of the job. A client no longer posts “type this PDF into Excel”. They post “manage my product catalogue across two stores and fix the data the AI tool gets wrong”, which is why Ecommerce Management was the fastest-growing admin skill at +130%. The same report puts AI Video Generation and Editing at +329%, AI Integration at +178%, and AI Data Annotation and Labeling at +154%. Every one of those is a former entry-level task with an AI tool inserted into the middle and a human responsible for the outcome.
If you were going to start with data entry, start with data operations instead: the same spreadsheet skill, applied to a client’s Shopify or WooCommerce store, where the deliverable is a clean, verified catalogue rather than typed rows.
Three beginner services that still pay in 2026
1. AI-assisted operations for small online stores
Product uploads, variant tables, image cleanup with an AI tool, description drafts that you then correct for accuracy and tone. You are selling reliability, not typing speed. The Upwork growth figure above is the proof clients are buying this. Learn one platform deeply (Shopify or WooCommerce) and one spreadsheet workflow before you write a single proposal.
2. Annotation and evaluation work, but only the skilled tier
Data labelling has split into two markets: bulk tagging that pays close to nothing, and domain-specific evaluation (medical, legal, Urdu-language, code review) that pays substantially more. We covered the divide in detail in AI training work has split into two tiers, and only one pays well. If you have any subject background, including a half-finished degree, that is your way into the better tier.
3. Video editing with AI tools in the pipeline
Short-form video editing for creators and small brands is still human-led, and the +329% growth in AI video generation and editing shows clients want someone who can run the tools, not just someone who owns Premiere. A beginner can build a five-clip portfolio in a fortnight using free tools and local businesses who need reels.
What to stop doing
Stop listing “content writing” as a standalone service unless you have a niche and proof. The 32% drop in basic writing and translation work is the clearest signal in the whole dataset. Writing is still bought, but as part of a bigger outcome: a landing page that converts, an email sequence with open-rate data, an SEO article that ranks. If you cannot show the outcome, the client will let a model draft it and skip you.
Stop chasing the Rs 500 to Rs 2,000 gigs on general platforms as a way to “build reviews”. Those gigs are where the shrinkage is concentrated, so competition for each one has gone up while the pay has gone down. A single Rs 25,000 project with a local clinic or a Dubai-based e-commerce seller does more for a profile than twenty rewrites.
And stop reading the export headline as a promise. Veqarul Islam of Jaffer Business Systems told the Tribune that part of the $1.6 billion freelance figure reflects salaried employees rebadged as freelancers for tax reasons. Whatever the true share, it is not new demand for beginners. We looked at what the data does and does not say about replacement in Will AI replace freelancers? What the 2026 data actually shows.
A 90-day plan for someone starting from zero
Weeks one to three: pick one of the three services above and learn the tool stack, not the theory. For store operations that means a free Shopify trial, a product CSV and an AI image tool. For annotation it means completing the qualification tests on two platforms and noting which domain questions you got right. For video it means editing five reels for a friend’s business.
Weeks four to six: do two projects for free or near-free for people you already know, and write up each one as a before-and-after with numbers the client agrees to. Those two case studies are your portfolio. A Fiverr gig with no proof is a lottery ticket.
Weeks seven to twelve: apply to no more than five jobs a day, each with a proposal that names the client’s actual problem and shows the closest case study. If you want structured training alongside this, the government’s PSEB SkillTech programmes are one route, and we broke down which of the six PSEB SkillTech programs is actually for you so you do not waste a month in the wrong one.
Where this leaves a beginner in Pakistan
The bottom rung is thinner, not gone. Under 9% of listings are still entry-level, but that is not zero, and the admin, research and transcription categories still appear in Upwork’s top-ten lists. What has been removed is the option to start with zero skill and let volume carry you. The beginner who wins in late 2026 chooses one narrow outcome, runs the AI tool a client would otherwise run themselves, and takes responsibility for getting it right. That responsibility is the part the chatbot cannot sell, and it is the part clients are still paying for.
