Silicon Valley has a lot to learn from the spreadsheet jockeys it hates & more related News Here

The most important question in financial markets can be put this way: Will companies like JPMorgan Chase and Walmart find artificial-intelligence models useful enough to buy them at a price high enough for OpenAI and Anthropic to stay in business? If they do: rock ‘n’ roll. If not, investors will tire of pouring capital into AI model-builders, destroying confidence in the massive construction of data centers and the semiconductors and gas turbines that power them, and so on to economic oblivion.

Last year the answer was clear, at least in Silicon Valley. The back-breaking labor of office workers in the 21st century will soon give way to the back-breaking work of the 20th century. That is to say, it will disappear, but this time in computers instead of offshore: companies will buy tokens, not workers. Tech bosses’ messiah-like belief in this issue has made them unpopular but not yet profitable. This partly reflects his confidence in technology. But it also reflects the office work scene that took shape in early 2020, when tech firms hired vast numbers of paper-pushers whose emails, dashboards, and coffee-chat “tech” employees despised. Since then the “bullshit jobs” approach has replaced “bullshit tasks” in the world.

Yet the war on white-collar jobs has progressed much more slowly than technocrats would like. The information that makes up modern companies is not easily organized and fed into AI models. The responses that models spew are not easily acted upon. For example, how should a firm decide which tasks to automate? When should it use cheap open-source models instead of expensive state-of-the-art models? How should it calculate returns on its investments? Could it reduce the cost of the token? Answering these questions is difficult and potentially very rewarding work. Hence the idea – simple, elegant and brilliantly ironic – that model-makers should be more involved in how their inventions are used. The bullish case works like this: Companies gain more from AI; Model-makers solve their existential cash flow problems; The world economy is progressing happily.

To destroy the advisor, then become the advisor. Each AI lab has announced partnerships with large professional-services firms. OpenAI recently purchased a small British tech consultancy. In May Anthropic and OpenAI both set up joint ventures with large private-equity houses, possibly hoping that AI models could replace many software companies that are losing their funding. That private equity cannot sell its ownership is an existential problem for the industry; If things remain this disappointing even after the attention of AI labs, it will be an existential problem for everyone.

Bodies should be thrown at the problem, but how many and by whom? When big tech companies built out their cloud-computing businesses in the 2010s, much of the work was outsourced to Accenture, a multinational consulting company that has grown enormously by applying the innovation of a handful of giants like Microsoft and Google to its client companies around the world. It has warm bodies galore, which it describes in the impenetrable language of the corporate central planner: 30,000 workers will be trained on the cloud in “a massive investment in talent, solutions and go-to-market strength,” it announced in December (a small contingent of Accenture’s total core of about 800,000 souls).

The market does not believe that Accenture will play the role of cash generation middleman this time. This year the price of its shares has declined by more than half. Maybe the investors are right. That said, ego, as well as transaction costs, are an underappreciated aspect of what companies want to bring in-house and leave to market: some at the top of AI labs think that a corporate culture that rewards individual genius is incompatible with hiring thousands of unskilled workers to incorporate their products into the economy. (Banks manage to combine star traders with huge back offices, although AI types may also find this analogy insulting.)

Palantir is close to eventually building on OpenAI and Anthropic — and that firm’s share price has also recently fallen on the idea. The essence of Palantir’s business is to send consultants into firms, sift through their data, and present it in a way that makes a middle manager feel as if he or she works at GCHQ. Apart from being an influential company, it is also an influential linguist. In its hands a firm’s data becomes its “ontology”. Consultants are given the status of “Forward-Deployed Engineers”. What differentiates them from Accenture’s new “redeployment-deployed engineers” or OpenAI’s “forward-deployed experts”? Who knows, but pity any consultant who remains backward and unemployed.

Which companies get to tell companies how to use AI models depends largely on whether model-makers develop a better understanding of the mind-numbing nuances of modern business than current consultants. Yet nothing AI companies have done so far indicates even the slightest understanding of how normal companies work. Perhaps this is not surprising for an industry whose employees pride themselves on being incompatible with bureaucratic corporate life.

ideal student

Whether Silicon Valley AI labs become white-collar consultants or not, there is a lot to learn from studying them. In investment banks, they will find case studies of being hated and dealing with systemic risks: worryingly, popular distrust is difficult to overcome and regulation usually only occurs after a disaster. Among elite law firms, they will see the exceptional longevity of small and focused firms. And in consultancy, there are lessons on the pains of growing too fast, as typical companies have felt in recent years, but also on the dangers of making bold predictions about the future of capitalism. This is not a warning they will heed.