Ando raises $20 million to rebuild workplace chat around AI agents

What Ando is building

Sarah Du arrived at the idea while helping companies build MCP servers in 2025. Clients repeatedly asked how to use AI agents directly inside Slack. The practical problems were familiar: messages had to move back and forth, and agents needed the right context without consuming an entire token budget.

Du’s conclusion was that the limitation was architectural. Slack and Teams were built for a world in which agents were applications people installed. Ando assumes they are becoming team members.

That changes the role of the human. Instead of forwarding an agent’s output to colleagues, employees can work in shared discussions where the agent can understand why a decision was made, ask questions, continue another agent’s work and bring in a person when judgment is needed.

Each agent in Ando has its own identity and inbox. It can participate in conversations alongside people rather than waiting for a direct command.

The product includes:

Agents can inspect channels and choose which ones to join. They can enter a conversation without being mentioned, and they can message an employee themselves when they decide that person needs to know something.

That is a more ambitious model than adding a bot to an existing chat product. It also creates a more difficult trust problem: an agent that can enter discussions and contact people has to make good decisions about when its involvement is useful.

The market Ando wants to change

Slack has turned its own built-in bot into an AI agent. Microsoft is integrating Copilot into Teams and connecting it to the rest of Office 365. Both companies already own the communication layer where many businesses coordinate work.

Ando’s argument is that a platform designed for agents can still outperform one adapting software built for people alone. The company is not the only new entrant making that bet. Jack Dorsey, the co-founder of Twitter and Block, recently introduced Buzz, a messenger that combines people and AI agents, although Buzz is more focused on developers.

The distinction Ando is pursuing is not simply that agents exist in the product. It is that they are expected to behave like members of the team: reading context, joining relevant conversations and coordinating people without being explicitly assigned every step.

The early test

The initial product was not an obvious success. People who saw early versions often found a less convenient messenger, according to Du. They formed that impression before reaching the part that made Ando different: the agents.

The signal changed when customers kept using the platform for longer. Ando showed that agents could spot the same issue being discussed in different channels, assemble the relevant participants in a group conversation, explain the context and suggest a solution.

Du sees this as evidence that agents may coordinate human interaction better than people can, because they can process far more messages in a short period.

Ando currently works with customers in three industries:

Those customers are spread across 15 countries, although many of the teams remain small. The new funding will support hiring and higher token spending.

The unresolved question

I think Ando’s most interesting claim is not that agents can answer questions. Slack, Teams and their AI features already cover that territory. The stronger claim is that agents can notice connections across conversations and manage parts of the coordination layer themselves.

That also exposes what the announcement leaves quiet about: how much supervision these agents require, and how often their unsolicited participation is genuinely useful rather than distracting. The early users are mostly small teams, so the product has not yet demonstrated that its model works when conversations, permissions and competing priorities become much harder to track.

Du believes agents could let very small teams operate at a scale that once required hundreds of employees. Agents would take on more execution, research and coordination, while people spend more time on evaluation, strategy and deciding what to do next.

If that happens, Ando will not merely be competing with Slack’s interface. It will be competing over who is responsible for keeping a company aligned—and whether that responsibility can be delegated without creating a new class of human go-betweens.