Google just announced a major expansion of its AI & Economy Research Program, and the roster reads like an economics department fantasy draft. Philippe Aghion—yes, the 2025 Nobel laureate—is joining as an Academic Advisor. Ajay Agrawal from Toronto's Rotman School is coming in as a Visiting Fellow. And they're bringing in two heavyweight research directors to actually run the thing.
This matters because Google isn't just studying AI adoption in the abstract. They're sitting on telemetry from billions of queries, Workspace usage logs, and enterprise deployment patterns. Now they're pairing that data exhaust with the economists who literally wrote the papers on technology diffusion and creative destruction.
The roster is stacked
Let's talk about who's actually joining. Aghion's work on innovation-led growth and creative destruction is foundational—he models how new technologies disrupt existing industries and what that means for long-term productivity. Having him advise on AI's macroeconomic trajectory is the right call, especially when everyone's still arguing about whether we're in a productivity boom or a vibes-driven hype cycle.
Ajay Agrawal holds the Geoffrey Taber Chair in Entrepreneurship and Innovation at Toronto. The announcement says he'll focus on "the economics of AI and scientific discovery, AI and robotics, and how AI can expand the frontier of human welfare." That last phrase is doing a lot of work, but Agrawal's research on prediction machines and the economics of AI gives him credibility here. He'll be working alongside David Autor from MIT, who's already a Visiting Fellow.
The two new research directors are where this gets operationally interesting:
-
Anu Madgavkar spent two decades at McKinsey Global Institute leading research on labor markets and technology adoption. She's advised governments and multilateral institutions, which means she knows how to translate econometric findings into policy frameworks that might actually get implemented.
-
Daniel Rock comes from Wharton and has co-authored foundational papers on technological transitions. He's an AI2050 Early Career Fellow and MIT Digital Fellow. The announcement says he'll "bridge frontier model telemetry with rigorous econometrics" to study enterprise productivity and labor restructuring.
That last bit is the key. Rock isn't just analyzing survey data—he's getting access to the actual usage logs from frontier models in production environments.
Why this structure matters
Madgavkar and Rock will work alongside Alex Imas (Director of AGI Economics at DeepMind) and Zanna Iscenko (AI & Economy Lead in Google's Chief Economist's Office). This isn't a ceremonial advisory board. It's a research org with dedicated leadership, external academic credibility, and internal access to the data.
The announcement frames this as part of the AI & Economy ATLAS program, which Google launched to track "how people are using Google's AI tools at work and in their daily lives." They're building an interactive site with global data on adoption patterns. But as they correctly note, "tracking adoption patterns is only the beginning."
The real question is: what happens after adoption? Do businesses actually get more productive, or do they just shift time from one low-value task to another? Do workers upskill, or do wages compress? Does innovation accelerate, or do we just get better-written emails?
The research agenda
Google says the program focuses on four core areas:
- Future of work – How AI changes job composition, task allocation, and skill requirements
- Productivity and growth – Whether AI actually moves the macroeconomic needle
- Global technology diffusion – Who adopts AI, how fast, and what barriers exist
- AI's impact on scientific discovery – Whether AI accelerates research itself
That fourth bucket is particularly interesting given DeepMind's AlphaFold and AlphaProof work. If AI can meaningfully accelerate scientific progress—not just automate grunt work but actually expand what's discoverable—that's a different economic story than "ChatGPT helps me write performance reviews faster."
The announcement emphasizes "rigorous economic inquiry" paired with "fine-grained data." That's the right framing. Survey data about whether people think they're more productive is noisy. Telemetry showing what tasks actually take less time, paired with econometric analysis of output quality, is the kind of evidence that might actually settle debates.
The conflicts are obvious but manageable
Let's address the elephant: Google has a vested interest in showing that AI drives productivity gains, upskills workers, and creates broadly shared prosperity. They're not a neutral observer.
But that doesn't make this work worthless. Academic advisors have reputational stakes—Aghion and Agrawal aren't going to sign their names to cooked numbers. The Visiting Fellows program brings in external economists who publish in peer-reviewed journals. And Google's Chief Economist's Office has historically published credible research, even when it complicates the company's narrative.
The real test will be whether they publish null results and negative findings. If every paper shows that Google's AI tools boost productivity by double digits, we'll know this is reputation management dressed up as research. If they publish messy, qualified findings—"productivity gains in X context but not Y, with Z confounders"—that's how you build credibility.
What to watch for
The announcement says they'll focus on "identifying the organizational practices, public policy frameworks, and training programs needed to ensure AI upskills workers, democratizes expertise, and drives broadly shared prosperity." That's aspirational language, but it points to the right questions.
We should see papers on:
- Complementarity vs. substitution: Do AI tools make workers more valuable, or do they make workers easier to replace? The answer likely varies by occupation, but the data should tell us where.
- Organizational adoption patterns: Which companies get productivity gains, and what do they do differently? Is it about complementary investments, change management, or just selecting the right use cases?
- Distributional effects: Who captures the value? If AI boosts productivity but wages don't rise, the gains accrue to capital. If wages rise but only for high-skill workers, inequality widens.
- Global diffusion dynamics: Does AI adoption follow the same patterns as prior technologies, or is cloud delivery and API access changing the diffusion curve?
The interactive ATLAS site should make it easier to track this work as it ships. If Google's serious about transparency, they'll publish datasets, replication code, and null results alongside the headline findings.
The bottom line
This is Google making a serious bet that rigorous economic research can help navigate AI's economic transition. They're hiring the right people, asking the right questions, and—at least rhetorically—committing to a multidisciplinary approach.
The proof will be in the papers. If this turns into a credible research program that publishes inconvenient truths alongside optimistic findings, it could meaningfully inform how we think about AI's economic impact. If it turns into a reputation-management exercise with cherry-picked results, the heavyweight roster won't save it.
Either way, having Philippe Aghion, Ajay Agrawal, and Daniel Rock working with frontier model telemetry is going to produce some interesting research. Whether it produces credible research depends on what they're willing to publish when the data doesn't cooperate with the narrative.