Chat, Notes, Channels, and Knowledge

SUPERWISE Chat is more than a box you type questions into. It’s a small set of surfaces, each shaped for a different kind of work — a quick question, a piece of writing you keep, a conversation a whole team is part of, a body of source material the AI can draw on. Once you know what each surface is for, you stop fighting the tool and start reaching for the right one without thinking.

This page explains the surfaces — Chat, Notes, Channels, Knowledge, Folders, and Assistants — and, just as importantly, when to use each. None of them is mandatory. You can do everything from a single chat window forever. But most people find that the moment a thread of work outgrows a single conversation, one of these surfaces is exactly the thing they were reaching for.

The shape of it, in one picture

Think of it as layers, from the most momentary to the most permanent:

Chat is where a single line of thinking happens — you and the AI, back and forth.
Notes are where a good answer goes when you want to keep it.
Channels are where a conversation stops being yours alone and becomes a team’s.
Knowledge is the source material the AI reads from before it answers.
Folders are the workspace that holds related chats, sources, and settings together.
Assistants are a saved setup — a job, a voice, and a set of trusted sources — that any chat can run against.

The first three are where the conversation lives. The last three are what shapes it. That’s the whole mental model. The rest of this page fills it in.

Chat — the default surface

A chat is one conversation between you and the AI. It’s the surface you land on, and for a lot of work it’s the only one you’ll ever need: ask a question, get an answer, follow up, move on.

Every chat is saved automatically and lives in your sidebar, newest first. You can rename it, pin the ones you keep coming back to, search inside a long one, and archive the ones you’re done with. Nothing is lost just because you closed the tab.

Reach for a plain chat when the work is yours, it’s a single thread of thinking, and nobody else needs to be in the room. A quick draft, a question you want explained, a problem you’re working through out loud — this is its home.

When a chat grows — when you keep asking the same kind of question, or you want it to stick to specific source material, or you want to bring a teammate in — that’s your signal to reach for one of the other surfaces. You don’t leave chat behind; you give it more to work with.

Notes — where good answers go to stay

A chat is a conversation. A note is something you decided was worth keeping.

The difference matters. Conversations are exploratory and a little messy — that’s their job. But every so often the AI produces something you’ll actually reuse: a clean summary, a paragraph of copy, a checklist, a table you’ll paste somewhere else. Leaving that buried in a chat means hunting for it later. Saving it as a note pulls it out into a clean, findable place.

You can:

Write a note from scratch — a free-standing piece of text you own and edit, with a title, a category, and a scope that controls who can see it.
Save part of an answer — highlight a useful passage in any response and save it straight to a note.

Notes keep a quiet link back to where they came from, so a saved answer always remembers the conversation that produced it. That provenance is the point: a note isn’t a disconnected snippet, it’s a kept result you can trace back to its source.

Reach for Notes when an answer has graduated from “interesting” to “I’ll use this.” Conversations are for thinking; notes are for keeping.

Channels — a conversation a team is part of

A regular chat is yours. A channel is a team’s conversation — a shared, persistent room where several people and the AI all talk together.

Channels are built for ongoing, multi-person work rather than a one-off exchange. They add the things a group conversation needs:

Threads — reply to a specific message in its own side-conversation, so the main channel stays readable.
Pinned messages — keep the important decisions and links at the top, where nobody has to scroll to find them.
A moderator — someone who can rename the channel, manage its members, and pin or unpin messages.

Everyone in a channel sees messages arrive in real time, and each person’s message is clearly attributed to them — so a teammate’s reply never looks like it came from you. You can address a specific assistant in the conversation using the $ picker in the composer, which is unique to channels.

There are two related ways to bring other people in:

A shared conversation lets you hand a single chat to specific teammates to read or collaborate on, in real time. It stays a conversation; it just has more people in it.
A direct message is a private conversation with one or more teammates, with the AI still in the room. A DM can be promoted to a full channel from its header when it outgrows a private thread.

Reach for a Channel when the conversation belongs to a group, not a person — a project standup, a shared investigation, a team that wants the AI as a participant. Reach for a shared conversation or DM when you just want to bring one or two people into something that’s still essentially a chat.

Knowledge — what the AI reads before it answers

Everything so far has been about where the conversation happens. Knowledge is about what the AI knows when it answers you.

By default, the AI answers from its general training. That’s fine for general questions. But when you need answers grounded in your material — your policies, your product docs, your contracts, last quarter’s report — you give it a Knowledge collection to read from. A collection is a curated set of documents the AI consults and cites in its answer, so you can see exactly where each claim came from.

You build a collection by:

Uploading files — PDFs, Word and Excel documents, slide decks, spreadsheets, text, and more (up to 100 MB each). Documents are read, indexed, and made searchable so the AI can find the relevant passage rather than the whole file.
Crawling a website — point Chat at a public web address and it pulls the pages into a collection in the background. (For your safety, it refuses to crawl private or internal addresses.)

Knowledge has a scope, and it’s worth understanding because it controls who can see what:

Scope	Who can see it
Personal	Only you — strictly private, even from administrators
Folder	The members of a folder (project) it’s shared with
Tenant	Everyone in your organization

A collection can be promoted up this chain — from personal, to a folder, to the whole organization — when material proves useful beyond you. Promotion only ever moves outward and each step is permission-gated, so nothing becomes more visible by accident.

When the AI answers using a collection, you’ll see numbered citations you can click to open the exact source — title, snippet, and a confidence indicator. If a collection has nothing relevant to your question, the AI tells you so and answers from general knowledge instead, rather than pretending the source said something it didn’t.

Reach for Knowledge when “answer from the internet” isn’t good enough and you need “answer from these documents, and show me where you got it.”

Folders — the workspace that holds it together

Once you’ve got several related chats, a collection of source material, and a preferred way of working, a folder (also shown as a Project) is the container that keeps them together.

A folder groups conversations, the knowledge they should draw on, and the settings they should follow. New chats you start while a folder is open automatically belong to it, so related work stays related without any filing. Folders have members with roles — the creator is the owner, and others can be added as admins, members, or viewers — which makes a folder a natural unit for a piece of work a small group shares.

Reach for a Folder when a single chat has grown into a body of work — a project, a client, an investigation — with its own sources, its own people, and its own way of doing things.

Assistants — a saved way of working

The other surfaces are places. An assistant is a setup.

When you keep asking the same kind of question, you end up re-explaining the same context every time: who you are, what tone you want, which documents to trust. An assistant saves all of that once. It bundles:

Instructions — what the assistant is for and how it should behave.
A persona — its tone, length, and formatting style.
Linked Knowledge — the collections it should answer from.
Guardrails — the trust and safety rules it works within.

After that, you just pick the assistant and start typing — it already knows the job. Your organization comes with a built-in assistant to start from, and you can build your own, either by filling in the fields or by describing what you want in a short conversation.

Reach for an Assistant when a kind of work is recurring enough to be worth setting up once and reusing — drafting in a house style, answering from a fixed set of policies, playing a consistent role for a team.

→ For a deeper walk-through, see Choosing the right assistant.

Putting it together

These surfaces are designed to nest, not compete:

You start a chat. It draws on Knowledge so the answers are grounded. When an answer is worth keeping, it becomes a Note. When the work belongs to a team, it moves into a Channel. When it’s grown into a project, a Folder holds it all together. And when the way you work becomes a pattern, you save it as an Assistant.

You’ll rarely set all of this up at once, and you never have to. Most people start with a plain chat and add the next surface the moment they feel the pull — a saved answer here, a shared room there, a folder when the desk gets cluttered. The tool is built to grow with the work, not ahead of it.

Where to go next

How Chat is different from other chatbots — why grounding, citations, and team surfaces change what you can trust the answers for.
Find and organize your conversations — the practical how-to for history, search, folders, and sharing.
Choosing the right assistant — when the default chat is enough, and when to build a setup.
Upload a file — get your first document into Knowledge.