We want to develop a framework grounded in organizational science instead of trusting someone's instinct.
As artificial intelligence quietly reshapes the rhythms of organizational life, a global competition co-launched by Stanford HAI and Google DeepMind has drawn over two hundred academic teams into a shared inquiry: not merely what AI can do, but what it does to us — to the way we coordinate, trust, and think together. The winning proposal, born from Stanford's own Graduate School of Business, seeks to decode the hidden grammar of teamwork itself, using the same machine learning architectures that power language models to map what makes collaboration succeed or fail. This is not simply a research prize; it is an institutional declaration that the transformation of work is too consequential to be left to technologists alone.
- AI is already inside the workplace, but no one has yet mapped the full human cost and consequence of its arrival — that gap is now being treated as a crisis of understanding.
- More than 200 teams from 156 universities competed to define the terms of that understanding, signaling that academia is racing to catch up with a transformation already underway.
- The winning researchers are building a 'large coordination model' — essentially teaching a machine to read the grammar of successful teamwork from emails, meetings, and shared documents.
- Four other finalist teams are attacking adjacent problems: filtering AI-generated ideas, measuring collective intelligence, unlocking siloed expertise, and analyzing team behavior in real time.
- Stanford is institutionalizing the effort with a new AI and Organizations Lab, while Google DeepMind is opening its own offices as a living laboratory for the winning research.
A question is becoming urgent inside organizations everywhere: as AI becomes a fixture rather than a novelty, what actually happens to the way people work together? Stanford's Human-Centered AI Institute and Google DeepMind designed a competition to surface the best thinking on that question, drawing proposals from over 200 academic teams across 156 universities. Researchers were invited to explore three broad territories — using AI to improve organizational alignment, understanding the human experience of AI deployment, and simulating how teams behave under different conditions.
After a double-blind review by judges from six universities and Google DeepMind, thirteen finalists pitched their work in person. The winner was a Stanford pair: PhD student Yankai Wang and organizational behavior professor Amir Goldberg, who proposed studying the hidden 'grammar' of coordination. Their insight is that teams coordinate through sequences of actions — emails, meetings, document edits — but no one yet understands why certain sequences succeed in some contexts and fail in others. They plan to build a transformer-based 'large coordination model' that learns from effective teams and predicts which coordination patterns will work best in a given situation. The prize includes $100,000 and the rare opportunity to test their model inside Google DeepMind's own offices, with full engineering and mentorship support.
Four other finalists were recognized at the May 12th conference, proposing work that ranged from applying lean manufacturing logic to AI-generated decision-making, to measuring collective intelligence, surfacing expertise trapped in organizational silos, and analyzing team behavior in real time using multimodal language models.
Beyond the competition itself, Stanford announced a new AI and Organizations Lab led by senior fellow Melissa Valentine, funded by Google DeepMind. The message from both institutions is deliberate: the transformation of work by AI is already happening, and it must be studied with human experience placed firmly at the center.
Over two hundred academic teams from universities across the globe submitted proposals to a competition designed to answer a question that's becoming urgent: as artificial intelligence seeps into the workplace, how will the way we work together actually change?
Stanford's Human-Centered Artificial Intelligence Institute partnered with Google DeepMind to launch the AI for Organizations Grand Challenge, an open call for researchers to imagine and study the future of collaboration in a world where AI is no longer a novelty but a fixture. The competition drew submissions from 156 universities. Teams of faculty and doctoral students proposed research across three broad areas: using AI to improve how organizations align around shared goals, understanding what happens to people when AI gets deployed in their workplace, and building simulated organizations to test how teams might behave under different conditions.
Melissa Valentine, a senior fellow at Stanford's HAI and an associate professor of management science, framed the moment plainly: "This competition marks the beginning of a broad, public conversation about how organizations are changing." The field has been moving faster than most people realize, and the challenge was designed to surface the best thinking about where it's headed.
A panel of judges from six leading universities and Google DeepMind evaluated the proposals in a double-blind process, looking for novelty, real-world impact, and whether the ideas could actually be executed. Thirteen teams made it to the final round, pitching their work in person. The winner came from inside Stanford itself: Yankai Wang, a PhD student at the Graduate School of Business, and Amir Goldberg, a professor of organizational behavior there, proposed to study what they call the "grammar" of coordination. Their insight was simple but powerful. Teams coordinate through a series of actions—emails sent, meetings held, documents edited—but nobody really understands what makes one sequence of actions effective in one situation and ineffective in another. Wang and Goldberg plan to use transformer-based machine learning, the same architecture that powers large language models, to build what they're calling a "large coordination model." The system would learn from how successful teams actually coordinate and then predict which sequence of actions will work best in a given scenario. The winning team receives one hundred thousand dollars and something more valuable: the chance to test their work inside Google DeepMind's own offices, with the company providing computing resources, engineering support, and mentorship.
Four other finalists were recognized at the competition's conference on May 12th. A team from Emory, Cornell, and Carnegie Mellon proposed applying lean manufacturing principles to organizational decision-making—helping leaders figure out which of the many ideas generated by AI are actually worth pursuing. Carnegie Mellon's Tepper School submitted a proposal to measure collective intelligence in teams using AI. Researchers from UC Berkeley's Haas School and INSEAD imagined using AI to surface expertise that's trapped in organizational silos. And Northwestern's Kellogg School proposed linking team science theory with multimodal language models that could analyze team behavior in real time to improve collaboration.
The competition is part of a larger institutional commitment. Stanford HAI announced the creation of a new AI and Organizations Lab, led by Valentine, funded by Google DeepMind. The message from both institutions is clear: the transformation of work by artificial intelligence is happening, and it needs to be studied carefully, with human experience at the center. As Simon Bouton, Google DeepMind's Chief Experience Officer, put it: "We look forward to continuing this collaboration to shape the future of AI in organizations."
Notable Quotes
We want to develop a framework to help leaders understand the dynamics of coordination and make decisions grounded in organizational science instead of having to trust someone's instinct.— Yankai Wang, PhD student at Stanford Graduate School of Business
This competition marks the beginning of a broad, public conversation about how organizations are changing.— Melissa Valentine, HAI senior fellow and associate professor at Stanford
The Hearth Conversation Another angle on the story
Why does it matter how teams coordinate? Isn't that just how work has always happened?
It matters because AI is changing the speed and scale of coordination. Right now, leaders make decisions about how teams should work based on intuition or what they've seen before. But if we can understand the actual patterns of what makes coordination work, we can help organizations adapt faster and smarter.
So the winning team is building a machine learning model that watches how teams communicate?
Exactly. They're using the same kind of architecture that powers language models, but training it on coordination patterns—the sequence of emails, meetings, document edits that lead to successful outcomes. It's like learning a grammar, but for how work actually gets done.
And then what? They just tell leaders what to do?
Not quite. The model predicts which sequence of actions will work best in a given scenario. It's giving leaders a framework grounded in data instead of just instinct. That's the shift—from "I think we should do this" to "here's what successful teams actually did in similar situations."
What about the other finalists? What were they trying to solve?
Different angles on the same problem. One team wants to help organizations decide which AI-generated ideas are worth the time and resources to pursue. Another wants to measure collective intelligence in teams. Another is trying to surface expertise that's hidden in organizational silos. They're all asking: how do we use AI to make collaboration better, not just faster?
Is there a risk that all this optimization makes work feel less human?
That's exactly why Stanford and Google DeepMind are funding a whole lab dedicated to this. The competition itself was framed around human-centered AI. The researchers aren't just asking "what can AI do?" They're asking "what should AI do in organizations, and what happens to people when it does?"