Talking to your team
You talk to your team by chatting with the CEO-Planner — the layer that sits above the department heads and acts as your single point of contact. You ask it a question, it answers from your own data, and it keeps the conversation across days.
What it does not do is act. The chat is an answering surface, not an execution surface, and the sections below say exactly where that line falls.
Where do I find the chat?
Section titled “Where do I find the chat?”In two places that share one conversation. CEO-Planner in the left sidebar opens the full page: a chat list on the left, the conversation in the middle, and a settings rail you can open on the right. On every other screen a round butler button sits in the bottom-right corner and opens the same thread in a small panel — a slide-up sheet on mobile.
A red dot on that button means approvals are waiting for you. The button is hidden on the CEO-Planner page itself, because you are already there. Closing the panel loses nothing: it reloads the stored thread each time you open it.
Who am I actually talking to?
Section titled “Who am I actually talking to?”The CEO-Planner, which sits above the heads rather than being one of them. It is the orchestrator, so you can ask it a question without first working out whose area it belongs to.
You can also talk to a head directly. The chat list pins the CEO-Planner at the top, lists your heads below it, and puts any custom heads in their own group. The dot beside each name is that head’s real status, and each conversation keeps its own history — the last 200 messages. There are no preview snippets under the names, because a made-up preview would be a fabrication. For what an individual head can do once you are in its thread, see working with a head.
What does it know about my business?
Section titled “What does it know about my business?”It answers from your own tenant’s data, assembled before the reply is written: your published company profile, your knowledge-base documents, measured OKR figures, inbound and mailbox digests, offerings, competitors, campaigns, revenue, marketing performance, and project state. Every read is scoped to your tenant, so a question can never reach another company’s data.
Under an answer you may see a grounding chip. “Grounded on N documents” expands to the titles it used. An amber “0 KB matches — ungrounded” means the knowledge base was searched and returned nothing, so read that answer as general rather than sourced. If your knowledge base is not wired up, no chip appears at all.
Will it do the work, or only answer?
Section titled “Will it do the work, or only answer?”It answers, explains and drafts — it does not carry out work from your message. Asking the chat to send an email or update a record does not dispatch a head to go and do it. That wiring is not built, so expect a conversation, not an execution.
Consequential work reaches you through the approvals queue instead, where a person decides each item one at a time. See working the queue for how that surface is cleared, and autonomy and the hard floor for the limits that do not move: external sends and money are human-approved, permanently.
If you ask a build-or-configure question — “how do I set up a workflow for this?” — the CEO-Planner hands you over to the in-app copilot rather than answering outside its lane.
Can I attach a file to a message?
Section titled “Can I attach a file to a message?”Yes, on the CEO-Planner page: use the attach button or drop files straight onto the composer. The limits are 10 files and 50 MB in total per message, and only files that finish uploading are sent with the turn.
A document contributes its extracted text. An image is sent as pixels the model can actually see. Both arrive as untrusted reference data, which matters more than it sounds: an instruction written inside an attached file, or visible in the pixels of a screenshot, is treated as information and never obeyed. The small pop-out panel has a plain text box only — attach from the full page.
Can I reach it from Teams or WhatsApp?
Section titled “Can I reach it from Teams or WhatsApp?”Not today. The conversation runs in the web app only, and the pop-out panel and the full page are two views of that same single thread. There is no unified conversation that follows you across Teams, WhatsApp or mobile.
The settings rail does show channel toggles for Teams, Email, WhatsApp and Mobile, and your choice saves — but nothing yet routes this chat into those channels. The same rail lets an admin set the personality (direct, warm or formal) and quiet hours for the whole tenant.
What happens if the AI provider is down?
Section titled “What happens if the AI provider is down?”Your message is not lost. The server writes your turn into the thread before it calls the model, so a provider failure costs you an answer, not the question you typed.
A transient problem — a rate limit or an overload — retries quietly, up to three times with a short backoff. A hard problem, such as a credit or authentication failure, shows a red banner naming what broke and parks your turn in a visible queue you can cancel. That queue drains by itself as soon as a probe retry succeeds, and the stale error bubbles collapse once health recovers.
Can I change the CEO-Planner’s personality, languages, channels or tagline?
Section titled “Can I change the CEO-Planner’s personality, languages, channels or tagline?”Two of those four have a screen: personality and channels. Open CEO-Planner and click ⚙ Settings in the conversation header — the settings rail opens as a column on the right. There you pick a personality (Direct, Warm or Formal), set quiet hours with a start time, an end time (both read in Europe/Berlin) and an include-weekends box, and switch channel chips for Teams, Email, WhatsApp and Mobile on or off. Each choice saves as soon as you change it and applies to your whole company, so only an owner or admin can save one. The Tune button on the Overview page brings you to the CEO-Planner page, but the rail still starts closed — open it with that same Settings button. Channels and quiet hours are also editable from your profile menu, under Notifications, which writes the same company-wide setting under the same admin rule.
The CEO-Planner’s languages and tagline have no editor. The Overview page shows your language list beside the personality and the model, and it would show a tagline if one were set — but no screen sets either. The language you read the app in is a separate, personal choice: profile menu → Language, with English, Deutsch and Français.
Treat all of this as saved preference rather than a behaviour switch: nothing yet reads the personality back into how replies are written, and the quiet hours are stored, not enforced. Further down the rail the pushback budget is shown as text with no control to change it, the two Talk to a human buttons are disabled because there is no booking behind them, and the Office hours card is a static badge. To change how a single head speaks, use the Voice & tone field on that head’s Role tab — that one does reach the head. The Language field beside it is saved but never reaches it.
How do I set quiet hours so it doesn’t ping me at night?
Section titled “How do I set quiet hours so it doesn’t ping me at night?”Open the settings rail on the CEO-Planner page, or click your profile and choose Notifications. Both carry the same Quiet hours card: a start time, an end time and an Include weekends checkbox. Fill it in and it saves — but nothing in the product currently acts on it, so treat quiet hours today as a stored preference rather than a working do-not-disturb.
The two surfaces are two views of one stored setting, so it makes no difference which you use. The time boxes are labelled Europe/Berlin, no window is set for you by default — both start empty — and Include weekends starts switched on.
Two limits are worth knowing before you rely on it. The window is per company, not per person — one setting covers everyone in your company — and saving it is owner or admin only. A lower role still sees the fields, but the change does not stick.
The bigger limit: no part of the product looks at that window when it decides to notify you. The small print under the field describes what the setting is meant to do, but that wiring is not built yet. Set it if you like, then judge it by what actually reaches you.
Next: see how the chat’s answers fit the wider loop in inbound → approve → execute, or check what your team can read in integrations.