We invested because humans& is pursuing a differentiated thesis at a critical inflection point: while today's frontier models excel at generating content, they often struggle to capture the collaborative dynamics that drive real enterprise value. 

humans& has assembled the rare team capable of solving this, combining deep research credentials with proven execution at scale.

 

Why this matters: Frontier models are approaching a data ceiling

The last wave of frontier progress was powered by scaling: larger models trained on vast corpora of internet text.

That approach produced systems that are fluent and increasingly capable. But it also created a constraint that is becoming harder to ignore: training primarily on static text is reaching diminishing returns for many real-world enterprise outcomes.

In organizations, value is created through interaction:

• People coordinate across threads and handoffs

• Teams negotiate constraints and tradeoffs

• Decisions evolve over time with new information

Today’s models can generate impressive outputs, but often struggle to reliably capture and improve the human collaboration layer that drives enterprise execution.

 

Why humans& is uniquely positioned to solve this

The challenge of building AI that understands human collaboration requires a rare combination: frontier research capability and proven execution at enterprise scale. humans& has both.

The founding team brings together researchers from OpenAI, Anthropic, Google DeepMind, Meta, and xAI — the organizations defining the current generation of frontier models. But critically, they also bring operational experience that demonstrates they can ship:

• Eric Zelikman (CEO) contributed to training data for Grok-2 at xAI and has published research on reasoning-focused reinforcement learning—precisely the technical foundation needed to train models on interactive rather than static data.

• Georges Harik (Co-founder) was Google's seventh employee, where he helped build Gmail, initiated Google Docs, and led the acquisition of Android—products that redefined collaboration and scaled to billions of users.

We invested because the problem they are solving — making AI understand how people actually work together — is both technically hard and economically significant. And humans& has the team capable of delivering on both dimensions.

 

Looking ahead: Building AI that augments human collaboration

humans& is preparing to launch its first product early this year, bringing this vision to market

We eagerly anticipate this launch because it will bring to life the conviction that we and the humans& team share: AI should augment human capability, not replace it. 

The most impactful systems will be those that learn from how people actually communicate and collaborate: capturing the nuance, context, and dynamics that make human teams effective. Humans& is building technology that can become more useful — both inside and outside the office — not by automating people away, but by understanding them better and helping teams work more effectively together.

We believe this approach will unlock a new category of enterprise AI: systems that don't just generate outputs, but that genuinely understand and enhance the way people think, decide, and create value together.

We are proud to partner with humans& in building this future.

Learn more at humansand.ai.