人工智能与人类：红色代码与程序员的终局

AI vs Human: Code Red and the Programmers' Endgame For most of the last year, we defended humans in the AI vs humans race - but this debatehas become more intense with obituaries of software programmers being written yet againwith every new AI tool launch. This, if true, would have far-reaching economic implications.Here, we focus on the realistic case and assess what’s likely in store. Venugopal Garre+65 6326 7643 Nikhil Arela+91 226 842 1482 Let’s start with the basics: are coders actually coding?Job titles can be misleading- they are coined by humans after all. Many with “manager” in their title aren’t managingpeople—they’re individual contributors. Countless Product managers aren’t setting productstrategy, instead spending much of their time managing backlogs. Similarly, the title“Software Engineer” is a misnomer. For numerous such engineers, bulk of their bandwidthis consumed not by coding, but operational overheads like deployment, management etc.In fact, actual coding takes up less than a fifth of their time. Today we boast of AI tools thatcan code better and faster than a typical software engineer, but are we truly replacing thatengineer—or just automating a fraction of his Job Description? Thepost‑hypereality:Since 2020, over a million tech workers have been fired acrossgeographies, with AI an (un)popular culprit. But are we there yet?: Enterprise adoption ofAI is limited - less than a third deploy any form of AI agents. At IT services firms, AI codingtools are in use, but we are yet to see clear signs of productivity gains. The acceleration inlayoffs is better explained by excess post-COVID hiring and weaker performance as overalltech services spending slowed. The sectors seeing layoffs in recent quarters are moreassociated with perceived AI impact than with the original COVID boom-bust (data now vstravel earlier, sales now vs fintech earlier), which creates the impression of an “AI effect” butis a combination of largely demand cycles and to a lesser extent firms eager to deliver theheadcount reduction metric with AI adoption, and have struggled with their decisions. The end game:The advent of AI also welcomes us to the era of wild extrapolations! Themarket believes a handful of AI platforms will replace most IT functions, capture the bulkof value, and employment in the sector will collapse. This narrative implies a dramatichollowing out in countries like India, where 10mn work in IT and BPM. But if one were tobelieve that complex systems like ERPs, core banking stacks, airline operations or decadesof code-base will eventually dance to AI’s tunes, and by extension most white-collar rolesfrom IT to trading and fund management are rendered redundant - why not take it to itslogical conclusion? Household incomes would fall, consumption would collapse, and wealthconcentration would be more skewed than ever. Why not then build in for a collapse in mostsectors, factor in large fiscal transfers and drive interest rates to absolute zero? In short,building in such a dramatic AI impact challenges all traditional portfolio constructs and is amacroeconomic catastrophe, not merely AI beneficiaries vs casualties debate. What we believe: The current market reactions perfectly capture two fallacies - theslippery slope and the bandwagon, and once the dust of the AI storm settles, we perceivea correction to emerge. Till then, statements like replacement of all white collar jobs willemerge and work wonders for valuations of AI centric companies. For the investors, suchperiods of heightened fear offer opportunities to buy beaten down sectors. DETAILS INVESTMENT IMPLICATIONS Technology is once again at a crossroads, with new business models emerging and uncertainty driving investors to extrapolateworst-case outcomes for incumbents. While the narrative often assumes rapid, disruptive AI adoption, the current realityis that enterprise usage remains limited, and even large tech companies and other enterprises are still pushing employeeshard to embrace AI tools, with uneven adoption and few clearly quantified productivity gains so far. Established developmentpractices, complex legacy code environments, and the need for deep contextual understanding create frictions that make arapid, wholesale break from current models less likely, at least in the medium term. Over time, AI is likely to be absorbed into thetoolbox of enterprise technology rather than replacing it entirely, becoming an important but ultimately incremental layer, similarto prior waves of automation rather than an immediate, all-encompassing substitute for human knowledge work. The extremescenario in which most knowledge jobs are eliminated appears, at this stage, more like a narrative extrapolation that just justifiescontinued heavy AI investment and encourages corporates to “adopt or be disrupted,” even when the realized benefits are stillmodest. In our view, the recent de-rating of AI-exposed or AI-adjacent sectors within technology may be somewh