Google hires a Nobel economist to figure out what AI is actually doing to jobs
Philippe Aghion, Ajay Agrawal and two new program directors join Google's AI & Economy team to track how the technology reshapes work, wages and small business.
Key points
- Google has added 2025 Nobel Laureate Philippe Aghion as an academic advisor to its AI & Economy Research Program, alongside University of Toronto professor Ajay Agrawal as a visiting fellow.
- Anu Madgavkar, formerly a partner at the McKinsey Global Institute, and Wharton economist Daniel Rock will run the program day to day as its new directors.
- The team will feed findings into Google's AI & Economy ATLAS, a public dashboard tracking how people use Google's AI tools at work and at home.
- The hires land the same week Google committed 13 billion euros to a Finnish data centre, its biggest ever European investment.
- Focus areas include the future of work, productivity, global technology diffusion and AI's effect on scientific discovery.
Google wants to know what its own products are doing to the economy, and it's hired some of the biggest names in the field to find out.
The expansion centres on what the company calls the AI & Economy Research Program, a small in-house group studying how artificial intelligence, the software behind tools like Gemini, is changing work and wages. The headline signing is Philippe Aghion, the French economist who won the 2025 Nobel Prize in Economics for his work on innovation-led growth. He joins as an academic advisor, sitting alongside Nobel Laureate Michael Spence and Cambridge's Dame Diane Coyle.
Ajay Agrawal, a University of Toronto professor and author of Prediction Machines, joins as a visiting fellow. He'll work alongside MIT economics head David Autor, already a fellow on the program.
Who is actually going to run this?
Two new directors. Anu Madgavkar spent two decades at the McKinsey Global Institute leading research on labour markets for governments and the United Nations. Daniel Rock arrives from the Wharton School at the University of Pennsylvania, where he co-wrote early papers examining which jobs generative AI can realistically perform.
Both will share the program with Alex Imas, who leads AGI Economics at Google DeepMind, and Zanna Iscenko from Google's Chief Economist's Office.
What is the ATLAS thing they keep mentioning?
ATLAS is Google's public dashboard, launched earlier this year, that tries to show in near real time how people use its AI tools. Think of it as a national statistics office, but for prompts instead of GDP. We first covered it on 20 July 2026, and our story from 17 September found that half of scientists now reach for AI every day, with adoption cutting across countries in ways that surprised even the researchers.
The new hires are meant to turn raw usage data into proper economic research: papers, models, forecasts. Areas on the list include small business adoption, enterprise productivity, scientific discovery and global technology diffusion.
| Name | Role | Coming from |
|---|---|---|
| Philippe Aghion | Academic Advisor | INSEAD, Collège de France |
| Ajay Agrawal | Visiting Fellow | University of Toronto |
| Anu Madgavkar | Program Director | McKinsey Global Institute |
| Daniel Rock | Program Director | Wharton, UPenn |
Why now?
Because Google's under pressure from every side to say, in numbers, whether its products help workers or replace them. Regulators want evidence. Chief executives want a business case. Unions want warnings.
The timing isn't subtle. Google committed 13 billion euros to a new Finnish data centre this month, its largest ever European investment, and published case studies on AI in disease, disasters and schooling. Meanwhile, new research we reported on 17 September shows wages falling faster in jobs most exposed to AI, even before those jobs disappear. Hiring Aghion's circle to scrutinise that dynamic is smart politics as much as it's smart science.
Here's the honest bit. A research program funded by the company whose tools it studies won't be treated as neutral, no matter how strong the names on the door. What it can do, if Madgavkar and Rock are given genuine room, is publish granular data nobody else holds: who is using AI at work, in which roles, with what effect on output. That would be genuinely useful. The test is whether the awkward findings get published alongside the flattering ones.
Google says the work feeds directly into future ATLAS updates. Watch what shows up there over the next year. That will tell you more than any hiring announcement.
