As AI becomes agentic, our jobs are bound to change
Artificial Intelligence (AI) is already speeding up everyday jobs like sorting invoices, answering basic customer questions, and writing first-draft reports. While some human tasks are disappearing, others are becoming easier because people are now using “co-pilots” that help them work more effectively and efficiently.
The next shake-up is about to occur with more powerful AI tools – reasoning LLMs – that will help leaders solve messy, high-stakes problems like where to invest, which products to launch, or how to reorganize a company. These tools will not just do number-crunching; they will reason, test ideas, and suggest actions.
Here’s the turning point: Instead of using general AI as is, companies will start transforming it, purposely designing or selecting agentic AI systems. These agents will embody theories and goals, learn from feedback, and run experiments to test those theories and reach those goals. Because companies will be able to custom design agents and feed them with their own data, rules, and know-how, agentic AI systems will behave like tailor-made strategists, tuned in to the firm’s view of the world.
For recurring management decisions, agents will watch, decide, and act on routine signals (like a spike in returns) and only call in the humans when something odd or sensitive pops up.
For more strategic and uncertain business decisions, agents will generate different scenarios for the firm’s future, run virtual tests, and guide leaders’ decision making. And while some clerical roles will fade, demand will grow for agent designers and orchestrators—people who build, test, and guide these AI systems, and who know when to trust them and when to pull the plug.
But in all this, there are two risks to be aware of. The first one is over-trust or mistrust: If leaders blindly follow an agent (or ignore it whenever they dislike its advice), decisions will be negatively impacted. The second risk is built-in bias: If an agent’s model or training data carry erroneous theories or prejudices, AI can repeat them at scale.
Success will favor organizations that can spot the right problems, define good theories to explain them, build or choose the right agents to tackle them, and govern those agents responsibly. Each time a company’s agent runs an experiment, it learns more about how the business world really works. That growing know-how is hard for rivals to copy.
In addition, we envisage the growth of a new division of labor, in which not only companies but also academics (particularly social scientists, business school scholars, and economists) can develop models of the world that train the agentic AI systems. Such an opportunity is also facilitated by the fact that LLMs are increasingly making it possible to code by providing these systems with instructions laid out in our own languages. This will be a critical function in that the ability to model the world is the outcome of the specialization of a particular category of knowledge workers, whom Adam Smith called “the philosophers and men of speculation . . . who are often capable of combining together the most distant and dissimilar objects.”
Summarizing, the first wave of AI made routine work faster. The next wave—agentic AI—will make companies think smarter and, to the extent they have the capability to design their own agents, in more unique ways. Firms that master these goal-driven AIs won’t just speed up old processes; they will discover new ways to compete and win.
The organizational ability to model problems in relatively general and abstract forms (theoretically, if we want to use the right word) will be critical. Along with this is the ability to feed and test these models with business data, and more generally with data that reflect the conditions of the problem at hand as closely as possible. In doing so, companies may have to rely to a greater extent on “philosophers and men of speculation” either by hiring them or by collaborating with the institutions that specialize in modeling and understanding the mechanisms that drive our problems and questions, whether in business, in our economies, or in societies at large.
Originally published in Fortune Italia
