IIf 2024 was the year of large language models (LLM), 2025 promises to be the year of AI “agents”. These are quasi-intelligent systems that leverage LLMs to go beyond their usual tricks of generating plausible text or responding to prompts. The idea is that an agent can be given a high-level – perhaps even vague – goal and break it down into a series of actionable steps. Once he “understands” the goal, he can develop a plan to achieve it, just like a human would.
OpenAI CFO Sarah Friar recently explained so at Financial Times: “It could be a researcher, a useful assistant for ordinary people, working moms like me. In 2025, we will see the first highly effective agents deployed to help people in their daily lives. Or it’s like having a digital assistant “who not only responds to your instructions, but is able to learn, adapt, and perhaps most importantly, take meaningful action to resolve problems on your behalf”. In other words, Miss Moneypenny on steroids.
So why are these automated Moneypennys suddenly considered the next big thing? Could this have something to do with the fact that the tech industry has spent billions of dollars building colossal LLMs with – as yet – no plausible return on that investment in sight? This is not to say that LLMs are useless; For people whose work involves language, they can be very useful. And computer programmers find them very useful. But for many industries, right now they still feel like a solution in search of a problem.
The arrival of AI agents could change that. Using LLMs as the building blocks of virtual agents capable of efficiently executing many of the complex task sequences that constitute “work” in organizations around the world could prove irresistible. At least that’s what the tech industry thinks. And so, of course, does McKinsey, the mega-consultant that provides the subliminal hymn that CEOs invariably sing along to. Agentic AI, stammers McKinsey“moves from thought to action” as “AI-enabled ‘agents’ that use basic models to execute complex, multi-step workflows in a digital world” are adopted.
If this is indeed what is going to happen, then we may need to rethink our assumptions about how AI will change the world. Right now, we’re mostly obsessed with the effects of technology on individuals or humanity (or both). But if McKinsey & Co are right, the most profound long-term impact could come from how AI agents change businesses – which, after all, are actually machines for managing complexity and transforming business. information into decisions.
The political scientist Henry Farrell, an astute observer of these phenomena, considered this possibility. LLM, he argues“are engines for summarizing and making useful large quantities of information”. Since information is the fuel that large businesses run on, they will embrace any technology that offers a smarter, more contextual way of managing data. information – as opposed to simple data that they currently process. So, Farrell says, companies will “deploy LLMs in ways that seem boring and technical except to those who are immediately involved for better or worse, but are actually important.” Large organizations shape our lives! As they change, our lives will change too, in a myriad of seemingly uninteresting but meaningful ways.
At one point in his essay, Farrell compares this “boring and technical” transformative impact of LLMs to the way the humble spreadsheet has reshaped large organizations. This caused a distinguished explosion by Dan Davies, economist and former stock market analyst whose book The irresponsibility machine was one of the nicest surprises of this year. He points out that spreadsheets “have enabled a whole new style of working for the financial industry in two ways.” First, it made it possible to create much larger and more detailed financial models, and therefore a different way of budgeting, developing business plans, evaluating investment options, etc. And second, technology has made it possible to work iteratively. “Rather than thinking about the most business-relevant assumptions and then projecting them, Excel [Microsoft’s spreadsheet product] encouraged you to simply lay out the predictions and then tweak the assumptions from top to bottom until you got an answer you could live with. Or, for that matter, an answer your boss could live with.
The moral of this story is clear. The spreadsheet was a revolutionary technology when it appeared in 1978, just as ChatGPT was in 2022. But it has become an integral part of organizational life. The advent of AI “agents” built from GPT-like models appears to follow a similar pattern. In turn, the organizations that absorbed them will also evolve. And then the world could possibly rediscover this famous adage attributed to John Culkin, colleague of Marshall McLuhan: “We shape our tools, then the tools shape us.”
What I read
Talk about economics
Transcription of a fascinating interview with the remarkable economist Ha-Joon Chang, on the economy, pluralism and democracy.
AI or A-no
“The false comfort of AI skepticism” is a vigorous test by Casey Newton on the two “sides” of the AI debates.
What Trump did next
“I have a cunning plan…” Charlie Stross’s blog is a sketch for a truly dystopian story on the consequences of Trump’s inauguration.
