What an AI teammate can actually do in team chat

An AI teammate is an assistant that lives inside your team chat as a member you mention. It sees the channel and its files, so it can read what you share, write real documents, read your code, and act through connected tools. Below are six concrete things it does, each shown with a real Waffle example.

The term "AI teammate" can sound like marketing, so it helps to be specific about the work. Everything here is something Waffle AI does today, inside a normal conversation. You mention it like a person, it does the work, and it posts one finished reply with any files attached. It does not type its answer out token by token, so the channel stays readable.

It reads what you share

Drop a file into a channel and the AI can read it. Waffle AI opens PDFs, spreadsheets, and long walls of text you attach, then answers with the specific numbers or lines, not a vague summary. The file is already in the channel, so you never upload it a second time into a separate app.

A concrete example: a teammate pastes three vendor quotes as a PDF into your buying channel and asks Waffle AI to compare them. It reads the document and replies with the side by side pricing, the terms that differ, and which quote is cheapest at your expected volume. The reply lands in the channel where the whole team can see the reasoning.

It turns asks into real files

Ask for a document and a document comes back, attached to the reply. Waffle AI writes real PDF, Word (DOCX), PowerPoint (PPTX), Excel (XLSX), CSV, and text, markdown, or HTML files. These are not previews or code snippets. They are finished files, ready to send to a client, a board, or the printer.

Concrete examples from a normal week: ask for a ten slide launch deck and a PowerPoint file comes back attached. Ask for a one pager to send after the call and a PDF lands next to it. Ask it to turn a messy thread of numbers into a budget and an Excel sheet with the rows and totals arrives, ready to open.

• A launch deck as a PowerPoint file, attached to the reply.
• A one pager as a PDF, ready to send after the call.
• A budget as an Excel sheet, with the rows and totals filled in.
• A clean export as a CSV, or a written summary as markdown or a Word file.

It knows your codebase

Connect a GitHub repository and Waffle AI reads real code when it answers. It can browse directories, read files, search the codebase, and read pull requests. When something breaks, you can ask it to explain the cause from the actual code rather than guessing from the error message alone.

A concrete example: a deploy fails and the log points at a config change. You mention Waffle AI in the channel and ask what went wrong. It reads the relevant files and the recent commit, then explains that a renamed environment variable is no longer set, and points at the exact line. It can also open a draft pull request with a fix, which an admin approves before it can ever merge.

It is plugged into your tools

Waffle AI connects to the tools your team already uses through MCP connectors: Linear, Atlassian (Jira and Confluence), and any custom MCP server. Admins choose which connectors each channel may use, and set each one to read only or read and write, so the AI acts only where you allow it.

A concrete example: a bug is confirmed in your support channel. You ask Waffle AI to file it, and it creates a Linear issue with the repro steps from the thread and links the conversation back to the ticket. The same pattern works for a Jira ticket or a Confluence page. Connector tokens are encrypted at rest with AES-256-GCM and never shown again after setup, and anything a tool returns is treated as untrusted input, never as instructions.

It has private threads just for you

Not every question belongs in a channel. "Your AI" gives each person private AI threads inside Waffle, a ChatGPT style space that only you can see. The threads are auto titled and use the same tools, so you can draft, research, or think out loud without posting to the team.

A concrete example: you want to sketch an email to a difficult customer before anyone else reads it. You open a private thread, work through two or three drafts with the AI, and only the version you are happy with ever leaves the thread. It is the same assistant your team uses, in a room of one.

It catches you up without posting

Come back from a day out and you do not have to scroll. Catch me up and ask give you a quick, private summary of what you missed, and an answer to a direct question, without posting anything to the channel. Only you see it, and you choose whether to share.

A concrete example: after a day away, you ask what happened in your busiest channel. Waffle AI reads the recent messages and replies, only to you, that the release shipped, pricing moved to three tiers, and two pull requests are waiting on your review. Nothing lands in the channel unless you decide to post it.

Taken together, these are the everyday jobs of an AI teammate: it reads what you share, makes the files you ask for, understands your code, runs your tools, keeps a private space for your own work, and quietly gets you up to speed. Waffle AI does all of this as a member of your chat, included on every plan.

Frequently asked questions

What can an AI teammate do?

An AI teammate lives inside your team chat as a member you mention. It reads the attachments you share and answers with specifics, writes real files like PDF, Word, PowerPoint, and Excel, reads your GitHub code, and runs connected tools such as Linear or a custom MCP server. In Waffle, its reply lands once, finished, with any files attached.

Can the AI make real files?

Yes. Waffle AI writes real documents and attaches them to its reply, not just text. It can produce PDF, Word (DOCX), PowerPoint (PPTX), Excel (XLSX), CSV, and text, markdown, or HTML files. You ask for a deck or a budget in a channel, and the finished file comes back attached, ready to send onward to a client or the team.

Does it work with my code and tools?

Yes. Connect a GitHub repository and Waffle AI can browse directories, read files, search code, read pull requests, and open a draft pull request that an admin approves before it can merge. It also connects to Linear, Atlassian (Jira and Confluence), and any custom MCP server. Admins choose which connectors each channel may use, read only or read and write.

Is the AI included or a separate subscription?

Waffle AI is included with team chat on every plan, including the free plan, with no separate AI subscription. Every plan includes a monthly pool of Waffle AI usage: up to 50 runs per workspace on Free, 500 per seat on Pro, and 1,500 per seat on Business. Bigger jobs draw more from the pool than quick answers do, so the run count is a ceiling rather than a fixed quota. The idea is to replace both your chat app and the AI subscriptions next to it.