Company Spotlight: Ando Builds Messaging for AI Teams

Ando is building team messaging for companies where AI agents participate in work instead of waiting in a separate window for another prompt. Founded in 2025 and headquartered in San Francisco, the company gives humans and agents a shared environment for channels, direct messages, group conversations, calls, context, permissions, and accountability.

Founder and CEO Sara Du previously co-founded Alloy Automation and became a Thiel Fellow after attending Harvard. Du started Ando after helping companies build MCP servers and repeatedly finding that agents could perform useful work but still depended on people to carry their output into the team conversations where decisions were made.

Ando matters now because AI capability is moving faster than organizational design. Companies can deploy agents across engineering, research, product, sales, and operations, yet the shared systems around those agents still assume every participant is human. Ando is trying to make coordination software catch up.

About Ando

Ando describes itself as team coordination software for humans and AI agents. The product includes familiar messaging features such as channels, DMs, threads, group conversations, video and audio calls, live transcription, searchable recordings, and workspace search. The difference is that agents receive their own identities, permissions, context, and place in the conversation.

In Ando's launch essay, Du explains that existing messaging systems bring agents in through infrastructure built for apps and integrations. Ando treats an agent as a workspace member that can decide which conversations matter, ask for missing information, connect related work, and hand judgment back to a person.

The company is agent-agnostic. Teams can connect external systems instead of adopting one model or one agent vendor, and Ando provides a hosted harness for customers that need it. The company currently charges by human seat rather than agent activity, an economic choice intended to keep teams from rationing useful machine participation by message or action volume.

Why Ando matters right now

The first phase of workplace AI concentrated on individual productivity. A person opened a tool, requested an output, reviewed it, and moved the result into the next system. That pattern becomes awkward when a company has several agents working across several functions. The model may be fast, but the organization still pays people to carry context between private sessions and shared work.

Ando's thesis is that the communication layer can absorb more of that coordination. An agent that was present for the decision can retain the relevant context, notice related work elsewhere, and participate in the handoff. The opportunity is less about adding another chat surface and more about reducing the distance between a decision, the work it creates, and the person or agent responsible for carrying it forward.

That creates a second responsibility: protecting human attention. Ando's product principles explicitly say that more proactive agents should not mean more noise. The system must decide whether an agent should speak, stay quiet, or keep watching, while respecting permissions and making its actions legible to the team.

Product design, memory, and security

Ando is building around persistent context rather than isolated prompts. Conversations can become memory for the workspace, allowing agents to understand why a decision was made instead of receiving only the latest request. The company also says customer context should remain portable and that its APIs and developer tools should make that context usable outside Ando.

The value of that memory depends on governance. Ando says agents cannot read a user's direct messages unless the context is forwarded. Its security program lists encryption in transit and at rest, role-based access, peer-reviewed production changes, monitoring, vulnerability management, and vendor review. Ando reports completing SOC 2 Type I in July 2026, with the SOC 2 Type II observation period in progress. The underlying assurance report is available to qualified customers and was not independently reviewed for this Spotlight.

Leadership, capital, and market position

Du's background gives Ando a founder-market connection rooted in integration and workflow design. Alloy Automation addressed the problem of software systems that could not easily work together. Ando moves that problem into a world where the participants themselves may be software, and where communication, memory, and permission sit inside the product rather than around it.

In September 2026, Ando disclosed $20M across pre-seed and seed financing. Accel led the pre-seed, while Index Ventures and Emergence Capital led the seed. The company reports use across 15 countries in software, finance, real estate, and professional services, although many current teams are small. Revenue, retention, customer count, and independently measured productivity gains remain undisclosed.

Slack and Microsoft Teams have distribution, integrations, administration, and enterprise trust that a new entrant must earn. Both can add agents. Ando's competitive argument is that the underlying assumptions matter: an agent that participates as a member may require a different architecture from an assistant attached to software originally designed for humans.

What Ando's hiring signals

Ando's current Research Engineer opening reveals where the company expects hard product work. The role focuses on evaluating proactivity, long-term memory, continual learning, permissions, redaction, and security using real, longitudinal workspace data. It also expects research to reach production within weeks rather than remain a benchmark exercise.

That hiring signal is more informative than a broad claim about growth. Ando is investing in the question at the center of its product: how to measure whether an agent's memory and intervention improve a team's work, including the cases where the correct action is silence. The company says it values evidence over credentials, strong technical communication, and comfort with incomplete, ambiguous real-world data.

Ando's public events add another layer to the company identity. Its 2026 program included discussions about language, consciousness, interface design, and human-computer collaboration, along with product previews and founder conversations. The pattern supports a culture interested in the social and philosophical consequences of agent participation, not only the mechanics of shipping another workplace tool.

The larger shift in team software

Workplace software has spent years helping humans organize information created by other humans. Agentic work changes the participant set. Identity, memory, permissions, escalation, and attention become core collaboration features because software can now observe, decide, and act inside the same environment as the team.

Ando is early, and its strongest claims still require market proof. The company must show that teams stay, that agents reduce coordination work, that proactive participation remains governable, and that the product can grow beyond small organizations without turning context into risk or conversation into noise.

The company's opportunity sits inside those obligations. If agents become ordinary coworkers, the winning interface may be the one that makes machine participation feel accountable, useful, and quiet enough to trust. Ando is building that layer while the definition of a team is still changing.