Most tech jobs in IT services can be automated with generative AI: Zerodha CTO

Zerodha CTO Kailash Nadh

Kailash Nadh defies the typical image of a chief technology officer (CTO) at a unicorn. Unlike other tech chiefs at major companies, he doesn't oversee hundreds of software engineers at Zerodha. In fact, his team of 30 coders has remained unchanged for the past couple of years, first through a bull market when techies changed jobs for doubling of salaries and then through a funding winter that saw thousands of them being laid off.

Nadh doesn't have to rush to release new products and features. This is because the stock trading platform takes its own sweet time in building and testing software before launching anything.

This leaves him with enough time to build open-source software and think about the world. Zerodha founder Nithin Kamath calls Nadh his "guru."

During hiring interviews, Nadh delves into personal and philosophical discussions with candidates, resulting in only two new hires making the cut in the last four years.

Recently, Kamath tweeted an internal note from Nadh that talked about the threat of artificial intelligence to jobs: "AI on its own won’t wake up and kill us all (for a while, at least!). The current capitalistic and economic systems will rapidly adopt AI, accelerating inequality and loss of human agency. That’s the immediate risk."

Meanwhile, the bootstrapped fintech has proposed an AI policy that promises not to lay off any employees because the technology makes a particular role redundant.

Subsequently, Moneycontrol had multiple conversations with Nadh, who incidentally completed his PhD in AI over a decade ago, to understand his views on how AI might impact society, its black box nature, the philosophical conundrums it presents, and more.

It really depends on what kind of software engineering we are talking about. When you are trying to build a sophisticated system that scales well, I don’t think today these tools can put a dent there. However, a lot of the lower-end technology jobs, where it's about taking an existing system

and configuring it slightly, or maintaining an existing system — which is most software work in the world — can be significantly impacted.

I have personally started using ChatGPT to debug software, whereas earlier I used to search for the output on Google and check what other developers have said about it. This is saving me 30-45 minutes per debug. Now, I get the answer in 5 seconds. That means one person is able to do many people's work. And the natural outcome of that is that you don't really need a lot of people.

Software teams used to be lean 10-15 years ago. Startups did not have the tendency to hire hundreds of programmers. However, with the venture capital boom, people have hired like crazy because the size of your team has become a growth metric. Normally, you don’t need large teams to write software. Perhaps, the increase in developer productivity due to AI tools will help reverse this trend. Or, maybe not.

It is the same 30 people. No changes in the last two years. In fact, we've only hired two people in the last four years.

The thing about us having a small team is that all 30 people handle a wide variety of things across systems. And they built a lot of domain knowledge. So, the density of work and the criticality of work each person does is more. But, imagine if we had 300 people. Then the density and quality of work that people would do would have been distributed evenly... their importance would be dissipated. I think it would be far easier to automate writing of software in large teams.

The reason behind IT services companies hiring a lot of people is not technology. It is about how their business is structured. The headcount is connected to billings in IT service companies. For example, a certain project could be billed as 30 ‘resources’ — that is the industry parlance — multiplied by the per hour charge. Although, it is not as if that money ever comes to the employees…It's a socio-economic reason why they hire so many and why they have such high attrition rates like 50-60 percent.

If you set aside all that, the actual technology work carried out by people at the lower ranks could easily be done by generative AI coding. There is very little creative work involved. It is all about maintenance and patching legacy systems for large clients. And, logically, that is the easiest thing to replace.

ALSO READ: Part 2 | 'OpenAI has a huge conflict of interest in calling for AI regulation'

It’s hard to say how it would play out. It's not easy for the replacement to happen because most of these large corporations have ancient technology, for which they can't find talented and creative programmers to maintain. People want to work on new and cool things. And all this work gets

offloaded to the back offices of IT services companies in countries like India, Indonesia, Philippines, etc.

But, it's just a matter of time. Because if automation becomes so cheap and so effective, then there will be economic pressures. Shareholders will start asking questions as to why are you paying 400 people, when an AI system can do it at a hundredth of the cost.

Goldman Sachs published a report recently saying that 300 million white collar jobs will be impacted. In the near term, that is a ginormous number. If many different kinds of white collar jobs can be obsoleted quickly, then is there a system in our global economic framework to absorb all the people who've lost jobs?

Our global existence is based on very specific socio-economic structures – education, job, retirement. Now, if that social order is disrupted by large scale job losses, I don’t know what that might mean. If we have to speculate, the middle class going away means that there'll be no world

left. The middle class is billions of people. They won't just go away, right? The social order will be upset. It's speculative. It's very dystopian. I mean, logically, it seems plausible. But who knows?

If you're speculating that the world is going to be out of order, because a huge number of people will be displaced, then I don't know if concepts like equity will have much meaning.

Only in a world where there is social order, the rule of law, and large markets, concepts like owning equity have any meaning. In a completely messed-up society, what does owning equity even mean? We can look at the countries over the last few years that have gone bust due to hyperinflation. If everything goes for a toss, when social order is lost, what is equity and what is safety?

This has been talked about in the context of AI in many science fiction books over the years, but it's pure speculation. It's probably even dangerous to speculate so extensively.

It is far more plausible than AI waking up and killing us all for sure. Even if it's not to such an extent, social disruption as a result of AI is far, far more likely than sentient AI trying to wipe out humanity.

If we really pushed ahead, I think that 20 percent could probably become 40-50 percent fairly easily. It can't be 100 percent for sure.

For example, we have a large number of people who handle physical forms, and who read physical forms, who take decisions based on the content and the physical forms that people send. If you start looking for avenues, then you can keep on automating very meaningfully, practically endlessly. Until you have very few people.

ALSO READ: Part 2 | 'OpenAI has a huge conflict of interest in calling for AI regulation'