I2024 was the year of large-scale language models (LLMs), and 2025 looks set to be the year of AI “agents.” These are semi-intelligent systems that leverage LLM to go beyond the usual tricks of generating plausible text and responding to prompts. The idea is that you can give your agent a rough (and sometimes vague) goal and break it down into a series of actionable steps. Once you “understand” your goals, you can create a plan to achieve them, just like humans do.
Sarah Friar, OpenAI’s chief financial officer, recently explained to the Financial Times: In 2025, we will see the introduction of the first highly successful agents that help people with their daily lives. ” or that we have digital assistants that “can not only respond to your commands, but also learn, adapt, and, perhaps most importantly, take meaningful actions to solve problems on your behalf.” It’s something. In other words, Miss Moneypenny on steroids.
So why are these automated money pennies suddenly being hailed as the next big thing? Even though the tech industry has spent trillions of dollars building huge LLMs, Could it have something to do with the fact that you still can’t expect a reasonable return on your investments? That doesn’t mean an LLM isn’t useful. This is extremely useful for people whose work involves languages. And for computer programmers, these are very useful. But for many industries, at the moment, they still seem like a solution looking for a problem.
With the advent of AI agents, things could change. LLM has the potential to be attractive as a building block for virtual agents that can efficiently perform many of the complex task sequences that make up the “work” of any organization. Or so the tech industry thinks. And, of course, McKinsey, the consulting giant that provides the subconscious hymn sheet every CEO sings. As “AI-enabled ‘agents’ that use underlying models to execute complex, multi-step workflows across the digital world” are adopted, Agent AI will move from “thinking to acting,” said Barbles McKinsey. “It’s on the rise,” he said.
Information is the fuel that runs businesses, and they will embrace any technology that provides a more intelligent way to process information.
If that really happens, we may need to rethink our assumptions about how AI will change the world. At the moment, we are primarily preoccupied with how technology impacts individuals or humanity (or both). But if McKinsey & Company’s claims are correct, deeper long-term effects could come through the way AI agents transform companies. After all, companies are actually machines for managing complexity and turning information into decisions.
Political scientist Henry Farrell, a keen observer of these issues, suggests this possibility. LLM, he argues, is “an engine for condensing vast amounts of information into something useful.” Because large companies are driven by information, they will adopt technologies that provide a more intelligent and contextual way of processing information, as opposed to just data, which they currently process. As a result, Farrell says, companies will “introduce LLMs in ways that seem boring and technical, except for things that are immediately relevant, for better or worse, but actually important.” Big organizations shape our lives! As people change, our lives will change in countless seemingly unexciting but important ways. ”
At one point in his essay, Farrell likens this “boring and technical” transformative impact of LLMs to the way a humble spreadsheet reshapes large organizations. This drew the genteel ire of economist and former stock analyst Dan Davis. His book, Machines Without Responsibility, was one of the year’s greatest surprises. He points out that spreadsheets have “enabled entirely new working styles for the financial industry in two ways.” First, it allows for the creation of larger and more detailed financial models, allowing for different ways of budgeting, creating business plans, evaluating investment options, etc. And second, this technology allows for repetitive work. “Instead of thinking about what assumptions make the most business sense and then sitting down and making predictions, Excel (Microsoft’s spreadsheet product) allows you to make predictions and sit back until you get the answer. I encouraged them to tweak their assumptions up and down. It was an answer that I could live with. What’s more, it was an answer that my boss could live with.”
The moral of the story is clear. Spreadsheets were as revolutionary a technology when they first appeared in 1978 as ChatGPT is in 2022. However, it has now become a routine and integral part of organizational life. The emergence of AI “agents” built from models like GPT appears to be following a similar pattern. In turn, the organizations that absorb them will also evolve. And in time, the world may rediscover the famous dictum of Marshall McLuhan’s colleague John Culkin: “We shape our tools, and our tools shape us.”
what i was reading
economics story
Transcript of an interesting interview with renowned economist Hajun Chan about economics, pluralism, and democracy.
Is it AI?
“The False Comfort of AI Skepticism” is a seminal essay by Casey Newton about the two “camps” in the AI debate.
Trump’s next move
“I have a cunning plan…” Charlie Stross’s blog post is a sketch of a true dystopian story about the aftermath of the Trump presidency.
