Field Guide · slack
Slack Is the Surface, Not the Tool
An agent that lives in a separate dashboard is a tool you remember to open. An agent that lives in Slack is a coworker your team forgets isn't human — which is the point.
The same question comes up at the end of every product walkthrough, after the agent profile and the constitution and the limits page: "so where does it actually live?" The honest answer is that the agent lives wherever your team already does the work, and for most teams reading this, that means Slack. The product decision to land in Slack first is not a UX preference. It is a bet about what makes an agent feel like a coworker rather than a piece of software you keep meaning to log into. Every Fidelic deployment ships into Slack on day one for that reason. Teams, email, and SMS are second-order surfaces — useful, eventually, but not the place the relationship gets built.
Why it matters
An agent in a separate web dashboard is a tool. You remember it the way you remember a reporting suite — when you need a number, you go look. You forget it the rest of the week. An agent in Slack is a coworker. It shows up in the channels the work happens in, gets pulled into threads the way a teammate gets pulled into threads, and answers in the rhythm the rest of your team answers in. The distinction is not cosmetic. Tools get used when there is time. Coworkers get used when there is work. Most knowledge teams have far more work than time, so the surface choice decides whether the agent earns its keep or quietly drifts to the bottom of someone's tab list.
Slack-native is not a checkbox; it is four distinct surface modes, each doing different work. A deployment that only uses one of them is leaving most of the value on the table.
Direct messages: the one-on-one channel
Every agent ships with a owner — the person on your team accountable for the agent's outputs and configured as its primary correspondent. The DM is where that one-on-one work happens: the morning briefing the agent has built overnight, the clarifying question the owner asks before the team standup, the mid-afternoon "draft me a response to this email" exchange. The owner reads the DM the way they read messages from a direct report, which is exactly the relationship the agent should sit in.
Channels: the team-visible deliverable
Channel posts are how the agent shows the rest of the team what it has been doing. The Monday-morning competitor digest, the weekly content calendar, the renewal-risk shortlist — these go in a channel where the team and the leader can see them, react to them, and forward them. The channel post does two things at once: it ships the deliverable, and it makes the agent's contribution legible to people who are not the owner. That second thing matters. An agent that only DMs its owner is invisible to everyone else, which makes it expensive to defend at the next budget cycle.
Threads: the handoff and the clarification
Threading is where the agent stops being a one-shot generator and starts being a teammate. When the agent posts a draft brief in a channel, the editor replies in-thread with "can you tighten the third paragraph and pull in the Q3 numbers?" The agent does, in-thread, without losing the original context. When a strategist asks a question in a thread that turns out to be partly an agent question and partly a human question, the agent answers the part it can answer and pings the right human for the rest. Threads are where the asynchronous handoff lives, and async handoff is most of how knowledge work actually moves.
App surface: slash commands and shortcuts
The Slack app surface — slash commands, message shortcuts, the Home tab — is for the structured asks that benefit from a form rather than a sentence. A slash command to spin up a new brief with a specified audience and length. A message shortcut to summarize a long thread the agent was not in. A Home tab that shows what the agent is working on this week. The app surface is the place where the agent looks like software, on purpose, because some asks deserve software's precision instead of a chat's ambiguity. It is the smallest of the four modes, but the one that makes the heavy-lift workflows repeatable.
The edge
There is a moment in the second or third week of a deployment that almost always tells you the surface choice was right. Two human teammates are in a thread arguing about a positioning line. One of them, mid-sentence, types "@" and tags the agent and asks for a tiebreaker draft. The agent answers in the same thread, in the same rhythm, and the conversation continues. Nobody pauses to context-switch. Nobody opens a tab. The agent has been pulled into a working conversation the way a fourth teammate would have been pulled in, except the fourth teammate is an agent, and the team did it without thinking. That is the test. The surface disappears. The work continues.
Honest take
This breaks for three kinds of company. Teams-first organizations where Slack is not approved for daily work — wait. Microsoft Teams support is on the roadmap; today it is not the deployment substrate. Regulated environments where Slack is blocked by IT or compliance — same answer. And companies where Slack exists but is treated as the chat app, separate from the work — the agent will land in a surface the team does not actually live in. The agent should not be the reason you adopt Slack. If your team is happily on Teams, wait for us, or skip.
So when the question comes up at the end of the walkthrough — "where does it actually live?" — the answer is the channels you already live in. Slack is the surface. The tool is the work.