AI and its labor market effects in the knowledge economy

In this episode, we discuss a new theoretical framework for understanding how AI integrates into the economy. We read the paper Artificial Intelligence and the Knowledge Economy (Ide & Talamas, JPE), and debate whether AI will function as a worker, a manager, or an expert. Read on to learn more about the model, our thoughts, timestamp, and at the end, you can spoil yourself on Andrey and Seth’s prior beliefs and posterior conclusions — Thanks to Abdullahi Hassan for compiling these notes to make this possible.

The Ide & Talamas Model

Our discussion was based on the paper Artificial Intelligence in the Knowledge Economy by Enrique Ide and Eduard Talamas. It is a theoretical model of organizational design in the age of AI. Here’s the basic setup:

* The Setting: A knowledge economy where firms’ central job is solving a continuous stream of problems.

* The Players: We have Workers (human or AI) and a higher-level Solver (human manager/expert or AI). Crucially, the human players are vertically differentiated—they have different skill levels.

* The Workflow: It’s a two-step process: A worker gets the first shot at solving the problem. If they fail, the problem gets escalated up the hierarchy to the Solver for a second attempt.

* The Core Question: Given this hierarchy, what’s the most efficient organizational arrangement as AI gets smarter? Do we pair human workers with an AI manager, or go for the AI worker/human manager combo?

* There are also possibilities not considered in the paper, such as chains of alternating managers and employees, something more network-y etc.

Key Debates & Critiques

Here are the most interesting points of agreement, disagreement, and analysis we wrestled with:

* Is a Solver Really a Manager? We spent a lot of time critiquing the paper’s terminology. The “manager” in this model is really an Expert who handles difficult exceptions. We argued that this role doesn’t capture the true human elements of management, like setting strategic direction, building team culture, or handling hiring/firing.

* My Desire vs. Societal Growth: Andrey confessed that while he personally wants an AI worker to handle all the tedious stuff (like coding and receipts), the economy might see better growth and reduced inequality from having AI experts and managers who can unlock new productivity at the highest levels.

* The Uber Driver Problem: We debate how to classify jobs like Uber driving. Is this already an example of AI managing the human (high-frequency algorithmic feedback), or is the driver still an entrepreneur who will manage their own fleet of smaller AI agents for administrative tasks?

Go Deeper

Check out the sources we discussed for a deeper dive:

* Main Paper: Artificial Intelligence and the Knowledge Economy (Ide & Talamas, JPE)

* Mentioned Research: Generative AI at Work (Brynjolfsson, Lee, & Raymond on AI in call centers)

Timestamps

* [00:00] Worker, Manager, or Expert?

* [00:06] Who manages the AI agents?

* [00:15] Will AI worsen inequality?

* [00:25] The Ide & Talamas model explained

* [00:40] Limitations and critiques

* [00:55] Posteriors: updated beliefs

The Bets: Priors & Predictions

We pinned down our initial beliefs on two key questions about the future impact of AI agents, the foundation of our “Justified Posteriors.”

Prediction 1: Will Managing AI Agents Become a Common Job? What percentage of U.S. workers will have “managing or creating teams of AI agents” as their main job within 5 years?

Prediction 2: Will LLM-based Agents Exacerbate Wage Polarization?

* Seth’s Prior: 25% chance it WILL exacerbate. Reasoning: Emerging evidence (like the call center study)

* Andre’s Prior: 55% chance it WILL exacerbate. Reasoning: Skeptical of short-term studies; believes historical technology trends favor high-skill workers who capture the largest gains.

Our Final Posteriors

Prediction 1: Will Managing AI Agents Become a Common Job?

The model slightly convinced Seth that the high-skill vertical differentiation story might be stronger than he initially believed, leading to a small increase in his posterior for exacerbation.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit empiricrafting.substack.com