Summary
Ando has emerged from stealth with a team messaging application that gives AI agents their own identities and inboxes. The company is positioning agents as participants in workplace conversations rather than applications that employees must call on and relay between colleagues.
A messaging platform designed around agents
AI agents are increasingly being introduced into workplace processes, but most team communication software still assumes that people are the only meaningful participants. Ando is taking aim at that assumption with a messaging application designed for human workers and AI agents to operate in the same collaborative environment.
The startup has emerged from stealth with a product that it describes as a potential replacement for Slack and other internal messaging platforms used by companies working with AI agents. Its central design choice is to give agents their own identities and inboxes. Instead of functioning only as applications installed inside a team workspace, agents can appear as distinct participants in conversations.
The idea addresses a practical problem that has become more visible as businesses experiment with agent-based workflows: employees often have to act as intermediaries. An agent produces an answer or completes part of a task, and a human then copies the result into a channel, carries follow-up questions back to the agent, and returns with another response. Ando’s model seeks to remove that relay step by placing the agent inside the conversation where the work is already being coordinated.
Moving beyond the “human proxy”
Sara Du, Ando’s founder, told TechCrunch that she encountered this challenge while helping companies build MCP servers in 2025. People wanted to use AI agents from within Slack, but moving messages and relevant context between the platform and the agent presented technical obstacles. The process could also consume substantial token budgets when agents were given too much information or had to repeatedly reconstruct the background of a discussion.
Du’s conclusion was that the problem was deeper than a missing integration. In her view, Slack and Teams were designed for a workplace in which software was primarily a tool used by people. Agents, meanwhile, were becoming participants in the work itself. That distinction matters because a participant needs more than an invocation button: it needs a recognizable identity, access to relevant context, and a place to receive and respond to work-related communication.
In the model Ando is proposing, an agent could understand why a decision was made, ask a colleague a question, build on the work of another agent, and involve a human when judgment is required. The intended benefit is not simply faster messaging. It is a different division of labor in which the agent remains present in the workflow instead of disappearing behind a human operator.
Familiar collaboration features, with a different set of users
The product includes channels, direct messages, and group conversations, according to the company. It also supports live calls that can be transcribed and viewed by an agent. That combination suggests an effort to make agent participation part of the ordinary communication layer rather than confining it to a separate chatbot interface.
For teams, the appeal of such an approach would be continuity. An agent could potentially follow a discussion in the same place as its human colleagues, rather than receiving isolated prompts through a separate system. Access to conversation history and call transcripts could give it more of the background needed to respond to a task or continue work that began elsewhere.
Yet placing agents directly into conversations creates requirements that are different from those of a conventional bot. Team members need to know whether a message was written by a person or generated by an agent. They may also need to understand what the agent is authorized to do, which sources it can access, and whether a response represents an automated process or a human-approved conclusion.
The public information available about Ando does not detail its permission model, review controls, or audit mechanisms. Those details will be important for organizations that handle confidential or sensitive information. A platform may make agent participation easy, but ease of access cannot substitute for clear boundaries around who can see a conversation and who can speak on behalf of a workflow.
Context is the difficult part of agent communication
The technical challenge behind Ando’s product is not merely sending messages from one system to another. Agents need the right context at the right time. A team workspace can contain long message histories, unresolved questions, repeated discussions, and informal comments that may or may not be relevant to a particular task.
Giving an agent the entire history can increase noise and processing costs. Providing only a small excerpt can omit the decision or constraint that explains what the team actually needs. The usefulness of an agent therefore depends on how the platform identifies relevant messages, preserves the relationship between a discussion and a task, and distinguishes settled decisions from open questions.
This is also where token budgets become a product concern. If an agent must repeatedly ingest large amounts of conversation to remain informed, the workflow may become inefficient. If the system compresses or filters too aggressively, the agent may produce responses that are technically coherent but disconnected from the team’s actual reasoning.
An agent-native communication platform may eventually compete on these context-management capabilities as much as on its user interface. However, the available reporting does not provide detailed information about Ando’s underlying architecture, context retrieval methods, commercial model, customer base, or performance. It is therefore too early to assess whether the product offers a measurable advantage over established workplace platforms.
Identity and accountability become central features
Giving an agent an independent identity is a useful starting point, but identity alone does not resolve accountability. In a human-and-agent workspace, participants may need to distinguish among an agent’s generated suggestion, a deterministic workflow result, and an action that has been reviewed or approved by a person.
The distinction is particularly important when a conversation involves confidential information, operational decisions, or an instruction that could affect other systems. The more naturally an agent participates, the greater the risk that people may assume it has the same authority or understanding as a human colleague. Clear labeling and access controls can help prevent that confusion, but the platform must make those controls visible without making collaboration cumbersome.
Human escalation is another important part of the model. Du described a system in which an agent could bring in a human when judgment is needed. That principle recognizes that agent participation should not mean removing people from decisions that require context, responsibility, or discretion. In practice, the usefulness of escalation will depend on whether the system can identify uncertainty and route the issue to the appropriate person.
A broader shift in workplace software
Ando’s launch reflects a broader question for enterprise software: should AI agents remain tools that people invoke, or should they become software entities with a continuing presence in the workplace? Traditional messaging platforms are organized around human accounts, human notifications, and human attention. An agent-oriented platform would need to manage a much larger population of automated participants without overwhelming users.
That creates a balance between availability and restraint. Agents need enough access to understand ongoing work, but not so much access that they absorb information unrelated to their tasks. They need to be able to ask questions and collaborate, but not to create an impression of authority they do not possess. They also need a clear record of what they saw, what they produced, and when a human took responsibility for the outcome.
Ando is approaching this shift through the communication layer, which is strategically significant because shared conversations are where context, coordination, and decisions converge. If the product can reduce the amount of manual relaying without obscuring agent identity or weakening governance, it could offer a different template for enterprise collaboration. If it cannot address those controls, direct agent participation may simply move existing integration problems into a more crowded communication space.
For now, Ando has presented a product direction rather than a fully demonstrated replacement for established platforms. Its launch highlights the limits of collaboration tools built for a human-only workplace and points toward a future in which agents may have inboxes, identities, and responsibilities of their own. Whether that future becomes a practical enterprise standard will depend less on how naturally agents can send messages than on how reliably organizations can manage their context, permissions, and accountability.
Source: link