Ando Emerges From Stealth With 20 Million Dollars To Make AI Agents Full Team Members

The specific failure mode Du set out to eliminate is what she calls the meat proxy problem, the pattern where a human employee has to sit between an AI agent's output and the rest of the team, manually relaying what the agent found, decided, or produced into a conversation the agent itself has no way to join. Ando's platform is designed to remove that human relay entirely. Agents on the platform get their own identity, their own inbox, and defined permissions, letting them browse channels, join relevant discussions on their own initiative, participate in threads and live conversations, and message specific colleagues directly when they determine a person needs to be looped in or asked a question.

Importantly, Ando does not build or require its own proprietary AI model. The platform is designed to be agent agnostic, meaning teams can connect the cloud agents and harnesses they are already using, including Codex, Claude, and Grokbot, rather than being forced to standardize on a single vendor's model to get agents working inside the platform. That positioning lets Ando compete on the coordination layer, identity, permissions, shared context, and conversational participation, rather than on which underlying model performs best, a distinction that may matter as the underlying agent market continues to shift quickly between providers.

Adoption so far has come from teams across roughly 15 countries, spanning software, real estate, and financial services, according to the company. Ando's own team has worked exclusively inside the platform since January, using it internally across engineering, research, product, operations, sales, and general company building, a dogfooding period the company points to as evidence the platform holds up under real operational pressure rather than only in demos. Access remains gated behind a waitlist at launch, and the company has said the platform currently works best for teams of up to around 30 people, a scale limitation that suggests Ando is still tuning performance and coordination quality before pushing toward larger organizations.

Accel partner Ivan Zhou framed the investment as a bet on a structural shift rather than an incremental feature, arguing that agents are not simply another capability layered into existing workplace software but are becoming participants inside companies themselves, comparable to how earlier computing shifts each produced their own new systems for coordinating work. Emergence Capital general partner Santi Subotovsky, whose firm has focused on productivity software for two decades, said he expects the way people work with each other and with AI agents to change fundamentally, and that what teams need is a platform that understands the context of their collaboration well enough to help determine which tasks are better suited to a person and which to an agent.

Ando enters a workplace messaging category where the two dominant incumbents are already moving to accommodate agents rather than standing still. Slack has converted its native bot into an AI agent, and Microsoft has built increasingly deep Copilot integration into Teams, including a Channel Agent feature currently in public preview that lets individual channels get their own agent drawing on that channel's conversation and meeting history. The distinction Du draws against both is one of posture rather than capability alone: Microsoft's Channel Agent largely waits to be asked, while Ando's agents are designed to act on their own initiative once granted the relevant permissions. Whether that more autonomous posture proves durable as incumbents narrow the gap, or whether it introduces coordination risks large organizations are not yet ready to accept, is likely to shape how quickly Ando can move beyond its current small team, waitlist gated launch toward broader enterprise adoption.