Want to kill grunt work with AI? Careful what you wish for

By Parmy Olson

Common wisdom about generative AI is that it will displace lots of professional jobs. The more nuanced take: It’ll take lots of entry-level jobs.

I’ve asked several businesses over the last few months how they’re using newfangled generative AI models from the likes of Microsoft Corp., OpenAI and Alphabet Inc.’s Google, and a common theme is automating grunt work — the kind you’d give to a new recruit, like conducting market research or writing reports. That might make sense on the surface, but as with every technology revolution there will be a price, one that both businesses and graduates will have to pay as AI raises the bar to starting a career.

One example of the trend comes from Man Group Plc. The British hedge fund has been using a large-language model to synthesize lots of unstructured market data from external sources into pithy paragraphs that are presented to its portfolio managers. That work is normally carried out manually by junior analysts. “Now they can work on higher-value stuff,” says Tim Mace, the firm’s managing director of data and machine learning. 

Much the same is happening at some of Wall Street’s biggest banks, who have been experimenting with AI tools that can instantly answer questions about publicly traded companies or generate reports and one-page presentations, work typically done by analysts who’ve just stepped onto the bottom rung of the investment-banking ladder. Top executives at Goldman Sachs Group Inc., Morgan Stanley and other banks have been debating how deep they can cut those incoming analyst classes, The New York Times reported in April.

A survey this year by the UK’s Institute for Public Policy Research found that low-ranking office work has a 16 per cent higher chance of being taken over by AI compared to jobs for more senior workers. “For a given job, entry-level positions are more at risk than those of more experienced professionals,” the report said. Little wonder that a frequent piece of advice for adopting ChatGPT is to treat it “like an intern.”

What happens when businesses automate entry-level roles? One outcome is lower costs, but there might be a tradeoff. Grunt work can be a valuable route to learning the tricks of the trade and building the intuition needed in a more senior role. Sure, it can be tedious for a junior analyst at an investment bank to build an Excel spreadsheet with 20 tabs about a stock, or a 50-page slide deck about a potential deal, but the process can also help them develop a better sense for valuing companies. What’s seen as a hazing ritual in some industries, and low-value work, can be useful in training the next generation of senior managers.

It also poses an uncomfortable challenge for college grads targeting professions like law and banking in the coming years. The New York Times goes as far as to say that investment banks will “nullify the need to hire thousands of new college graduates.” If future employers expect their young recruits to do higher-level work, those grads will need to show they have those higher-level skills. Over the next one or two decades, that might even lead to a rethink of how higher-education prepares young people to enter a workforce where they are no longer doing the job of interns but managing machines that are interns instead.  

Technology has of course been streamlining entry-level tasks throughout history, taking us from abacuses to calculators, typewriters to word processors, and library stacks to smartphone databases. But generative AI represents a bigger leap: It's not only automating routine tasks but mimicking human creativity and decision-making, threatening to erase entire categories of junior positions and shutting established entry points to professional careers.

Businesses should be mindful of how they define “low-value” work that’s suddenly ripe for automation. It might be more valuable than they think.