Slackbot Can Now Draft Full Documents. Here Is What It Reads To Do It

Slack used to search your workspace and summarize what it found. As of September 2026, Slackbot writes the document itself. Big Mode is a full-screen workspace where Slackbot researches across your channels and connected apps, then drafts a multi-page brief, post-mortem, or status report you can push straight into a channel.

That is a genuine time saver for the first draft of something long. It is also a different kind of tool than a search box, because the words that end up in front of your team were assembled by a model reading things you did not hand it one at a time. Before you post a Big Mode draft under your name, it is worth knowing exactly what fed it.

What Big Mode Actually Pulls In

Slack’s own announcement describes Big Mode as giving Slackbot “room to do deep research, multi-step analysis, and real drafting, all without leaving Slack.” The companion feature, Deep Research, is explicit about scope: it digs across your organization’s data and the web, then synthesizes what it finds into a cited report. In Slack’s own example, a single report pulled from Slack channels, a connected Salesforce record, and outside web research on a prospect’s vendors.

That is three different trust levels blended into one draft: things your team wrote, things a connected app recorded, and things a stranger published on the internet. The draft does not visually separate them. A citation link tells you where a claim came from if you click it, but the paragraph reads the same whether the source was your own incident channel or a random blog post about your vendor.

The Exclusions Are Opt-Out, Not Opt-In

Slack gives admins real controls here, and they are worth setting up before Big Mode gets used on anything sensitive, not after. Admins and workspace owners can mark specific channels, canvases, and lists as excluded from Slack AI features entirely, so their contents are not used for summarization, search, or as context in a drafted report. There is also an organization-level toggle to turn the feature off outright.

The shape of the rollout follows the same pattern as most workplace AI features shipping in 2026: switched on for the workspace by default, with the burden on an admin to carve out exceptions. If your legal, HR, or incident-response channels have not been explicitly excluded, assume Slackbot’s research can read them the next time someone runs Deep Research on an adjacent topic.

Skill Sets Make This Easier To Roll Out Team-Wide

The third piece of the update, Skill Sets, lets a workspace bundle up to 25 related Slackbot skills and share them with a team in one click instead of one at a time. That is a convenience feature, but it has a side effect worth naming: once a skill set that includes Deep Research is shared to a channel, everyone in it can generate a cross-channel report without configuring anything themselves. The exclusion list, not the individual user’s judgment, becomes the only thing standing between a quick report and a document that quotes a channel nobody meant to include.

Check The Draft Before You Send It, Not After

The failure mode is not that Big Mode gets facts wrong. It is that a hedge from the research stage survives into a message that reads as settled.

Before:

Per the research Slackbot pulled together across the incident channel, support tickets, and a couple of related threads, it looks like the checkout outage was probably tied to the 2pm deploy, though the timeline in the incident channel and what support saw do not fully line up yet, so we are still confirming exactly what happened and when it actually resolved.

After:

The checkout outage ran from 2:04 to 2:41 PM, caused by the 2:00 PM deploy. Incident channel and support timelines now agree. Full write-up is in the pinned doc.

The first version is Slackbot’s research stage showing through: several sources, not yet reconciled, hedged accordingly. The second is what you post once you have actually confirmed the timeline yourself. A drafting tool that reads three sources at once is genuinely useful for assembling a first pass; it is not a substitute for the step where a person decides the draft is now true.

This is the same discipline covered in how Outlook’s Copilot draft-and-grade split works and in what Gmail’s AI Overviews read to answer a question in your inbox: a tool that reads more of your workspace to write something is doing you a favor at the drafting stage and handing you a verification job at the sending stage. Skipping the second half is where these features get you into trouble.

A Wrivio Context for tightening a Big Mode draft before you post it could say:

Rewrite this as a short status update for the team. Keep every date, number, and root cause exactly as written. Remove hedge phrases like “it looks like” or “probably” only where I have confirmed the claim myself; otherwise flag them so I catch what still needs verifying. Cut anything restating the source report rather than stating the conclusion. No em dashes.

Press Ctrl+Shift+Space, paste the draft, and check the diff. A rewrite that keeps an unconfirmed hedge is doing its job by leaving evidence you still need to check; one that quietly smooths every hedge into a flat statement has done you a disservice.

For channels you would rather Slackbot never read into a report at all, the underlying question is the same one covered in should you give an AI agent access to your inbox: access granted for convenience is access you now have to actively manage, not access that quietly stays out of the way.

Common Questions

What is Slack’s Big Mode?

Big Mode is a full-screen Slackbot workspace, launched in September 2026, for long-form research and drafting inside Slack. It lets Slackbot produce a multi-page document, such as a strategy brief or post-mortem, that you can then push into a channel.

What sources can Slackbot’s Deep Research pull from?

Deep Research draws on your organization’s Slack channels, data from connected apps such as Salesforce, and the open web, then returns a cited report. The citations link to the original source, but the draft text itself does not visually separate internal from external claims.

Can admins stop Slackbot from reading certain channels?

Yes. Admins and workspace owners can mark specific channels, canvases, and lists as excluded from Slack AI features, and there is a separate organization-level toggle to disable the feature entirely. Both are opt-out controls an admin has to set up, not a default state, so an unexcluded channel is fair game for Deep Research until someone configures the exclusion.

Does a Skill Set change who can run Deep Research?

Skill Sets let a workspace bundle up to 25 Slackbot skills and share them to a team in one click. If Deep Research is part of a shared skill set, anyone in that channel can generate a cross-channel report without setting anything up themselves, which makes the channel exclusion list more important, not less.

Should I post a Big Mode draft without editing it?

No. Treat a Deep Research report the way you would treat a colleague’s rough notes: useful for assembling the facts, not yet something you have personally verified. Confirm the specifics against the primary source before the draft goes out under your name.

Download Wrivio for Windows to tighten a workspace-AI draft into a message you have actually verified before it goes to the team.