AI will not kill jobs. It will kill the people who don't use AI

From a 6,000-rupee filing job to a million-dollar ARR

The episode opens with Muzamil welcoming Zeeshan Sikandar onto Thought Behind Things and admitting he has been meaning to have this conversation for four or five months. The framing is deliberate. Muzamil has been spending more of his time on the Pakistani tech industry because, in his words, “Pakistan’s future is now entirely in the hands of tech” — and he wants people who are actually building inside the industry, not commentators outside it.

Zeeshan’s origin story sets the tone. His father, a banker for the last thirty years, bought him a computer in second grade in 2002 and made him type out of Microsoft Word books every day. By ninth grade, when Facebook arrived in Pakistan, he taught himself phishing well enough to take over his entire class’s accounts. “It was a phishing attack,” he says, almost casually. That curiosity carried him into computers as a career, but the path through it was anything but smooth.

He could not afford college regularly. He took a job at a travel consultancy in Karachi, Pulani, starting at 6,000 rupees a month doing filing work. He automated the filing system — it used to take two hours to retrieve a file — and was moved up to selling international tickets, taking home eight to ten thousand rupees a month. Some of that money went home. The household was in a state Zeeshan describes plainly: “We were not able to afford chicken. We could not even afford eggs. Maybe once a month we’d have chicken or meat. That was the time we lived through.”

Muzamil pauses the story there. He wants to know whether Zeeshan ever felt sorry for himself, whether he ever sat down inside the constraint. The answer is the most quotable line in the opening. “I was always a dreamer. I am still a dreamer. The customer sitting in front of me — I had this hope that within ten years, within twenty years, I would be the one sitting in that seat. And that time was not too far.”

Why entry into freelancing closed after COVID

Zeeshan eventually finished his degree at DHA Suffa University in computer science, worked at HBL as a product development engineer on what is now a ten-million-user mobile app, and quit nine months in because the work was maintenance and there was no room to grow. He resigned with no active clients. The first Fiverr client came the next month. By 2019 the agency had started.

When Muzamil asks how a newcomer should think about freelancing today, Zeeshan is unambiguous. The market that he walked into in 2019 — where he ranked on the first page of Fiverr within two months — does not exist anymore. “Before COVID, entry into freelancing was easy,” he says. “After COVID, a remote culture set in. Millions of people moved onto these platforms. The market got very saturated. If I had to recommend it to someone new today, I don’t think they’d succeed the same way.”

The only durable opening, in his read, is to be early in a niche the platforms have not yet absorbed. Mobile apps in 2019. NFTs and blockchain in 2022. Generative AI in 2023. The agency retrained its resources each time. The current opening, he argues, is the one most agencies are still naming wrong — what used to be called SaaS, now AI-as-a-service. The number of gigs in generative AI and AI agents on Fiverr is still small. That is the wedge.

How AI rewrites every node of the agency

Muzamil walks Zeeshan through the agency lifecycle node by node — inbound lead, pre-sales, proposal, close, build, delivery — and asks where AI is actually showing up. The answer is granular enough to be worth following in detail.

Pre-AI, a proposal took three to four days of research, feasibility, and tech mapping. Post-AI, the same proposal is sent in an hour. The goal is to close on the first call before the lead cools. The ERD that used to sit in a junior engineer’s head for a week is generated in ChatGPT or Grok, which Zeeshan frames in a striking way: “We have an engineer with two years of experience building a backend. AI has a hundred years of experience across thousands of engineers. The foundation it builds is better.”

Development time is being compressed similarly. Cursor and GitHub Copilot are in active use across the team — purchased, not pirated. Work that used to take four and a half months can now ship in one. The pre-sales, sales, business-development, and design teams have all been pulled into the same workflow; UI Wizards generates inspiration that designers then build on. Even HR has been automated: a link is sent to the candidate, an AI agent conducts the first interview, and a hire/no-hire report comes back.

Muzamil pushes harder on the headline claim. What does “using AI” actually mean? Zeeshan’s framing here is the one that gives this episode its title. “AI is not going to kill jobs. AI is going to kill the people who are not using AI.” The person who is not using AI, he says, gets replaced. The person who learns to drive the tools — to escalate from ChatGPT to Claude to Cursor when one fails — keeps their seat.

The hidden cost: developers are losing the foundation

Zeeshan does not pretend this transition is free. When Muzamil asks about productivity gains among the engineering team, the answer comes in two parts.

As a CEO he sees only upside. A four-month project now ships in two; that frees the agency to take on a second project in the saved time, which is how the business has been able to expand. As an engineering manager, he sees a problem. “Our developers have become so reliant on these tools that the critical thinking they used to do has stopped.” Two months before the recording, the team ran a small no-AI test. The basics had decayed. The engineers had become very good at integrating AI-generated code, very good at switching between tools when one stalled, and noticeably worse at building logic from scratch.

His response inside the agency is to schedule deliberate no-AI sessions — “brain exercise,” as he calls it — so the logic-building muscle does not atrophy completely. The broader LinkedIn debate on vibe coding and whether LeetCode-style logic still matters runs through this section. Zeeshan’s read is that both sides are partially right. You cannot ignore AI. You also cannot lose the foundation underneath it.

Communication is the actual gap, not coding

Muzamil names something he says ninety-nine percent of his industry conversations land on: the number-one problem with Pakistani tech talent is not technical. The boot camps have solved that. The problem is communication. Zeeshan agrees without hedging.

“This is the biggest gap we work with,” he says. “An engineer might have done great technical work, but they cannot convey what they have done. You are very good in tech, but if you are unable to convey, there is no benefit to your tech.” He extends it. Asian engineers, in his experience, will hit a technical block and rather than surface it, they will make an excuse to avoid the conversation. That is what creates the perception abroad that Asian developers go silent — not that they cannot do the work, but that they cannot narrate it.

The structural fix inside Zeeshan’s agency is twofold. Only project managers and technical leads talk to clients. Everyone else writes their messages through ChatGPT first to improve clarity and context before sending. This is not optional — it is policy. The augmentation model, where every engineer is in front of the client, is something he has consciously avoided for exactly this reason.

The work-ethic problem underneath the communication problem

Muzamil presses Zeeshan on whether the issue is really just articulation or whether it goes deeper — to professional ethic, discipline, commitment. Zeeshan takes the broader frame.

“We are not even aware that what we are doing is wrong,” he says. He gives examples. An eleven o’clock meeting time is treated as one o’clock. A daily progress update the client expects at end-of-day simply does not get sent. “And the person doesn’t realise it is wrong. We have normalised it. We do the wrong thing with so much confidence that we have made it the right thing.” A European client recently sent Zeeshan a picture of a German office where phones go into a designated box and calls are taken outside; the contrast with the Pakistani office, where two hours of TikTok during work is treated as background noise, is the kind of thing Zeeshan brings up not to moralise but to mark how far the floor is from the developed-economy default.

He extends it to attitude. Fifty percent of the engineers his team interviews fail, in his estimate, on attitude alone. Two years into a career, the message becomes “we don’t do junior-level coding anymore.” The argument, in Zeeshan’s read, is not really an argument — it is a culture that has stopped associating effort with value.

Muzamil’s framing of the same problem is sharper. “Pakistanis are chasing shortcuts. They’re chasing Ponzi schemes. There’s an obsession with easy money. In the older world, money was money and work was work.” The mismatch is value. A junior engineer sees a friend earning four lakhs and assumes the salary is the entitlement, not the output. Zeeshan agrees: “He doesn’t know what value his friend is creating in that specific organisation. To get to that level, you need time. You need effort. You need to create the same value.”

The hard ceiling at Fiverr, and why San Francisco is the next move

Muzamil asks the question that determines the agency’s next decade. Can the vision — becoming one of Pakistan’s largest software services companies, on the trajectory of Tenpearls or Systems Limited — be reached through Fiverr alone? Zeeshan is direct: no.

The unit economics give it away. The agency’s average ticket size sits in the $40,000 to $50,000 range. Software houses doing equivalent work through direct relationships are at half a million to a million. Cold outreach, email marketing, PPC — all of it caps somewhere between twenty and a hundred thousand dollars per deal. Above that, the work has to be closed in a room. Tenpearls has offices in Columbia and North America. That is what Zeeshan is now building toward.

He had landed in San Francisco the Sunday before the recording. “I had never thought it before — when you land there, you realise you are nothing. You are doing nothing.” The agencies that get to ten million in annual revenue, he argues, are the agencies with a US back office, a US sales office, US-based hires, and US-side bank accounts. The Pakistan operation runs the engineering. The US operation closes the work.

When Muzamil asks whether Pakistan still has a cost advantage over India, Zeeshan picks his spot carefully. The structural advantage is the zero-tax regime — Pakistan does not tax IT exports or freelancers, India does. That is real and it is the floor. Beyond it, India has scale, depth, and a market the Western buyer trusts by default. Zeeshan’s own teams already include Indian freelancers. The point is not to compete on price; the point is to be present where the buying happens.

How Zeeshan reads the next twenty-five years

Muzamil closes with the question he asks every guest. Where is Pakistan in twenty-five years? Zeeshan’s answer is more grounded than optimistic.

The advantage, he says, is demographic. Pakistan’s population is overwhelmingly young, and the business culture is finally shifting away from the forty-plus seth-ji model toward young founders building something from scratch. He expects a correction over the next twenty-five years that leaves the economy materially better than it is today — conditional on political stability and on the government continuing to leave the IT sector alone.

The earlier part of his answer on AI carries the same temperament. Every industrial revolution kills jobs and opens new ones. Agriculture gave way to industry. Industry gave way to software. AI is the next iteration of the same pattern, and the people who get caught out are the ones who refuse to retrain. “If you are not using AI in your work, if you are not changing yourself,” he says, “the same thing will happen that happened during COVID. Traditional businesses will close. Your job will go. The same pattern repeats.”

By the end of the conversation, the through-line is hard to miss. The trajectory from a 6,000-rupee filing desk to a million-dollar ARR was not built on talent in the abstract. It was built on adopting the next thing two years before the rest of the market, communicating cleanly enough to keep clients, and being willing to move — physically, to San Francisco — when the ceiling of the current model became visible. Muzamil and Zeeshan land in the same place: the next decade rewards the people who actually use the tools, in the rooms where the buying happens.