Ando: Building the Interface Layer for Human and Agent Teams

I first met Sara last summer, seated next to her at a founder dinner we hosted in San Francisco. What struck me was her sharp, articulate ambition: reimagining messaging - the place where real work happens, where decisions get made, and where expertise gets shared - for a world where agents are teammates, not plugins. Around the same time, my partner Sahir met her separately and came away just as impressed. So we sat down with her over coffee to hear the full vision and by the time we left, we knew we wanted to invest.

Sara has been building towards this idea for a decade. She started coding in high school in Atlanta, then dropped out of Harvard at 19 to start her first business, Alloy. She is drawn to products and design as a way of understanding the world, and she thinks like a product designer first, inspired by the minimalism of her favorite architect, Tadao Ando.

That design sensibility matters more here than it might seem. Many AI-native tools built for this moment are optimizing almost entirely for what agents need, treating the human interface as an afterthought. But the winning messaging platform still has to be something people want to open every day, and will be judged on how intuitive and delightful it feels to use, not just how capable and optimized the models underneath it are.

Sara’s bet is that the messaging medium itself has to be rearchitected, with agents treated as first-class participants. Nothing in the organization is as valuable and yet so disposable at the same time as messaging. Bolting an AI assistant onto that mess doesn’t fix it - it adds another cook into an already crowded kitchen.

Ando is built around three layers: An agent execution layer, governing how agents perceive and act inside conversations, handling memory, permissions and the ability to extend across other apps. A human messaging layer, rethinking composition itself: writing a new message and searching for an answer become the same action, with AI reasoning applied before a human is ever pinged. And a memory layer, tagging every message at creation with intent, stage, topic and context, functioning as the retrieval engine underneath everything else.

Team messaging hasn't been meaningfully reinvented for years. Yet, the category is enormous, with Microsoft Teams generating close to $10 billion, built almost entirely for human-to-human collaboration. Now, we need a platform genuinely built for agents and humans at once. The platform that captures structured context first, tasks, decisions, and agent activity in one place, has a real shot at becoming the connective tissue for how hybrid organizations operate, with third party agents and integrations compounding around it over time into a wider ecosystem.

We believe Sara’s combination of ambition, discipline and polish is rare, and we’re very excited to be leading her Seed.