A new entrant with heavyweight pedigree is betting that AI’s next leap isn’t smarter answers but smarter teamwork. Humans& has secured a $480 million seed round to build a foundation model optimized for coordination, positioning its system as a “central nervous system” for how people and AI work together across projects, teams, and time.
Why Coordination Is The Next AI Battleground
Chatbots excel at one-off tasks: answering questions, summarizing documents, writing code. Real work, however, is collaborative and messy—priorities shift, decisions sprawl, and alignment drifts. That “many-people over long horizons” problem is where traditional assistants break down.
Investors are pouring money into the coordination layer for a reason. McKinsey’s 2023 analysis estimated generative AI could add $2.6T to $4.4T in annual economic value, much of it tied to knowledge work where coordination costs are high. Microsoft’s 2024 Work Trend Index reported that 75% of knowledge workers already use AI at work, but most usage remains siloed, not organizational. As Reid Hoffman has argued, the next gains come from reengineering workflows—how teams share context, sequence tasks, and make decisions—rather than sprinkling agents into isolated pilots.
Inside Humans&’s Approach to Building Coordination AI
Humans& was founded by alumni of Anthropic, Meta, OpenAI, xAI, and Google DeepMind. Their thesis: today’s models are optimized for correctness and immediate user satisfaction, not the social intelligence required to facilitate groups, negotiate trade-offs, and steward long-running commitments. The company says it is building a model and product in tandem to co-evolve behaviors, memory, and interface.
Technically, the effort leans on long-horizon reinforcement learning to plan, act, and revise over weeks or months, and multi-agent RL to operate in environments with multiple humans and AIs in the loop. There is precedent that this matters: Meta AI’s CICERO achieved human-level play in Diplomacy by combining strategic planning with natural language negotiation, while Stanford’s Smallville simulation showed emergent social behaviors among simple agents with memory. Humans& aims to productize those ideas for practical teamwork.
Crucially, the startup is not building a plug-in for existing suites. It wants to own the collaboration layer itself, potentially replacing multiplayer contexts like chat, docs, and task boards with an AI-moderated space that understands roles, goals, and norms. The model is being trained to ask value-aware questions—more like a colleague who knows when it’s worth interrupting—while maintaining persistent memory about users, teams, and past decisions.
What This Could Look Like in Real-World Practice
Imagine a product launch where marketing, sales, legal, and engineering disagree on timelines. Instead of a cascade of meetings and Slack threads, Humans& envisions an AI facilitator that gathers positions, highlights conflicts, proposes trade-offs, and locks decisions with audit trails. It could track commitments over quarters, ping the right stakeholders when assumptions break, and nudge teams back into alignment.
In a smaller setting—even a household—the same system might coordinate calendars, budgets, and preferences, learning how individuals like to be asked and when to escalate. To work, it must integrate with the tools people live in today while preserving privacy, role-based access, and compliance. Expect governance features, decision logs, and fine-grained controls to be as important as raw model capability.
A Crowded Field With High Stakes for Coordination AI
The opportunity is drawing competitors fast. Anthropic is pushing team workflows via Claude Cowork, Google’s Gemini is embedded across Workspace, and OpenAI has been pitching multi-agent orchestration and developer workflows. On the startup side, note-taking app Granola recently raised $43 million at a $250 million valuation to expand collaborative features.
Humans& faces steep risks. Training and serving a new foundation model requires massive compute and data, and the majors are not standing still. Incumbent collaboration platforms already sit where work happens; displacing them demands not just a better model but a superior day-to-day experience. There is also the classic platform trap: if coordination becomes a killer feature, larger players could fast-follow or pursue acquisition.
The Case For A Socially Intelligent Model
Still, the bet has a logic. Most productivity losses stem from coordination failures—unclear owners, missing context, churned decisions—problems ill-suited to single-turn chat. A model trained for memory, negotiation, and long-range planning could shrink that drag. If it works, we may see a new category: AI-native coordination systems that sit above apps, not inside them, and measure success in aligned outcomes rather than prompt satisfaction.
The technical hurdles are formidable: reward design for social outcomes, collecting data that reflects real organizational dynamics, and preventing manipulation or bias in group settings. But the upside—more coherent teams, fewer redundant meetings, and tighter execution—keeps drawing capital. With $480 million and a team forged at top labs, Humans& now has a runway to test whether coordination, not just cognition, is where AI’s next breakthrough will land.
