Betting Slack was built for the wrong workforce
Ando launched publicly on September 24 with $20 million in pre-seed and seed funding from Accel, Index Ventures and Emergence, positioning itself as a full replacement for Slack or Microsoft Teams rather than a plugin bolted onto either one. The pitch is specific: existing team messaging platforms were designed for human-to-human conversation, and every attempt to bring AI agents into them treats the agent as an installed app rather than a participant with its own presence in the conversation.
Founder Sara Du built the product around a problem she saw repeatedly as agents became more common in real workflows: agents finish work, but a human still has to relay that work into the channels and threads where the rest of the team actually operates. Du calls those humans 'meat proxies,' a blunt name for a role that exists purely because the messaging layer was not built to let an agent post, join, or start a conversation on its own.
What agent-native actually means here
In Ando, agents get their own identities and inboxes, the same as a human employee would. They can browse channels and join conversations without being explicitly tagged first, and Du says agents have shown the ability to notice when two separate conversations are converging on the same problem and proactively bring the relevant people together, explain the shared context, and even suggest a decision, all without being asked to intervene. The platform also supports live calls with transcription, group conversations, and direct messages across both human and agent participants.
The design decision to give agents first-class presence rather than app-integration status is the part worth examining closely. A Slack bot today can post messages and respond to slash commands, but it cannot join a conversation it was not invited to, or initiate a group chat on its own judgment. Ando is explicitly building for a workplace where an agent's judgment about when to interject is treated the same way a human colleague's judgment would be, which is a meaningfully different trust model than most enterprises currently apply to bots.
The real problem it targets
Du's stated design goals point to genuine friction that grows worse as organizations deploy more agents: the manual effort of relaying agent output between systems and people, the inefficiency of agents lacking shared context across conversations, and token budget concerns that come from having agents constantly re-explain what they already know. Any engineering leader running more than one or two agents in production has likely felt at least one of these firsthand, usually in the form of a Slack channel that has become an unreadable feed of bot notifications nobody actually reads.
That specific failure mode, the bot-notification channel everyone mutes, is arguably the biggest tell that current messaging tools were not built with agents as first-class participants. If your organization's agents currently post into a dedicated channel that has quietly become background noise, that is direct evidence of the exact gap Ando is targeting, and a sign that agent output is not actually reaching the people who need to act on it.
How far this has actually gotten
Ando's current customer base spans software, real estate and finance companies across 15 countries, but the announcement is explicit that these are primarily small teams rather than large enterprises. That is a normal and sensible place for a category-defining collaboration tool to start, but it also means the hardest test, whether agent-native messaging holds up inside a large organization with complex permissioning, compliance requirements and thousands of employees, has not happened yet.
Enterprise messaging is also one of the stickiest software categories that exists, with switching costs measured in years of institutional workflow habit, integrations and compliance sign-off, not months. Slack and Teams both know agents are coming to their platforms too, and both have shipped their own agent integration features already. Ando's bet is that retrofitting agent-native features onto a human-first architecture will always feel secondary, while a platform built agent-native from day one will feel native to both kinds of participants as agent adoption scales.
Why this is a real architecture question, not a UI preference
It is tempting to file this under interface design, but the underlying question is closer to identity and access management than to chat UX. Giving an agent its own inbox and the ability to initiate conversations means giving it a persistent identity, a permission scope, and an audit trail, the same governance questions that apply to any autonomous system acting inside your infrastructure. A messaging platform that treats agent identity as a first-class concept from the start has already made decisions about those questions that a bolted-on integration usually has not.
That is worth evaluating on its own terms even if you never adopt Ando specifically. Ask your current collaboration vendor how it handles agent identity, permission scoping and audit logging today, and compare that answer to what a purpose-built agent-native platform offers. The gap between those two answers will tell you how much technical debt your organization is accumulating every time it adds another agent into a chat tool that was never designed to host one.
What belongs on your roadmap
Do not wait for your incumbent messaging vendor to solve agent-native collaboration before deciding how your own agents should communicate with your team. Map out today how many of your production agents currently rely on a human to relay their output into the tools your team actually reads, and treat every instance of that as a workflow cost worth quantifying, not an acceptable permanent state. That mapping exercise alone usually surfaces more manual relay work than engineering leaders expect once they actually go looking for it.
Ando's arrival, regardless of whether it becomes the eventual standard, confirms that the collaboration layer has become a genuine competitive battleground for agent-heavy organizations rather than a solved problem sitting quietly in the background. Put agent identity, permissioning and communication architecture on your platform team's roadmap for next year alongside the model selection and orchestration decisions that have gotten most of the attention so far, because the tool your agents use to talk to your humans is quietly becoming as consequential as the model powering the agents themselves.
