- An assistant lives in WhatsApp, Slack, or Discord. It answers quick questions there and starts a brand chat thread for longer marketing work.
- A slash command sends a dashboard chat request to a specific specialist.
Review your assistant’s work
Open an assistant’s Working on list to see each job’s status and its latest recorded step. Select a job to read the request, result, next step, and execution history without leaving the assistant page. Open results takes you to the brand thread and its saved artifacts. Open thread lets you steer ongoing work through the normal chat controls. A job that finishes without all the requested work shows Finished with gaps. Review the thread for what remains. A running job’s heartbeat does not count as a completed step.Save completed work as a skill
Select a successfully completed job in Working on, then choose Save as skill. This sends a request into the same brand thread and uses a normal chat turn. The assistant’s orchestrator uses the available brief, your corrections, and the result to draft a repeatable procedure. If it lacks a step, it can ask you to clarify. The draft covers when to use it, required inputs, steps, and approval boundaries. It is scoped to the source brand and records the source thread and run in a reference file. Review and edit it in Skills before publishing it. Saving a draft does not make it active or grant permission to publish or send anything. Archive an unwanted draft from Skills. You can also tell your assistant, “Save how we did that as a skill.” It continues the thread that did the work. A failed or incomplete job must be finished first.Permissions and pause
In the assistant’s Settings > Instructions, Permissions shows the brand’s enforced defaults for connected-app changes, publishing or sending, and deletion. Connection restrictions and action-specific approval rules also apply. Owners and admins can inspect standing grants for the brand and assistant here, including provider, scope, expiry, and limits. Manage permissions and grants opens brand settings, where grants can be revoked. Permissions apply across dashboard chat, hosted channels, and recurring work. Instructions and memories guide how the assistant writes and plans. They do not grant access or bypass an approval. Use the permission settings for actions that must require approval. An assistant owner who is not a workspace admin can edit their assistant’s preferences but cannot change brand permission policy. Pause stops new assistant replies and recurring task starts. Existing brand threads continue; queued deliveries and approval notices can still arrive. Stop an active run in its brand thread if you want that work stopped too. Turn the assistant on to resume new work, and resend requests that arrived while it was paused.Set up an assistant
Each assistant works for one brand, talks on one channel, and has its own name, tone, and avatar. There are two kinds, and the channel decides which:
Open Assistants from the brand switcher. The page lists two slots for the brand you are in, a personal assistant (WhatsApp) and a team assistant (Slack or Discord). Choose Create on an empty slot, then give the assistant a name (it starts as “‘s assistant”), a tone, and an avatar. Casual chats short and warm, lowercase is fine, and uses emoji sparingly; Professional is crisp and direct. Creating it opens that kind’s channel connect straight away: WhatsApp pairing for a personal assistant, the Slack or Discord install for a team one. You never pick a brand, because an assistant works only in the brand whose page you created it from.
How many assistants you can have is the same on every plan:
The per-workspace count includes AI apps: each app that is on and working in a brand takes one workspace slot, like an assistant, but none of a brand’s personal or team slots. The meter at the top of the page counts both. When every workspace slot is in use, the empty slots say No free slot. Delete an assistant or turn off an AI app to free one, or contact us if you need more. A paused assistant still holds its slot. Assistants in your other brands are listed under In other brands.
To change the name, tone, or avatar later, open Settings on the assistant’s page and choose Identity. The name saves when you leave the field or press Enter; Escape cancels the edit. Tone and avatar changes save when you choose them. Standing instructions, the rules it follows in every reply and piece of work, are under Settings > Instructions. Instructions you gave it in chat (“from now on, always cc Sam”) are listed under the editor, where you can edit or remove each one.
Under Identity > Look, the avatar can be one of the presets, your brand’s logo, or a picture of your own. Choose the upload button to add an image; it is saved to the brand’s asset library. Choose the generate button to open a new brand chat with a prompt already written, such as a rounded version of your logo that works at small sizes. Edit the prompt (for example, “a girl version of my logo”) and send it. Up to three images that chat makes are added to Look with a dot marking them as new. Pick one to use it. To remove a picture from Look, hover over it and choose the x. The picture stays in your asset library.
A team assistant’s name and avatar also show in Slack, on each message it posts, and in Discord, as the bot’s nickname and picture in your server; if your Slack workspace connected before this was available, reconnect Slack to show them.
To remove an assistant, choose Delete assistant beside its pause button. Workspace admins can delete it after confirming. This disconnects its channel, unlinks its members, cancels queued messages, and stops its tasks and watches. Brand threads and drafts stay in your workspace. The assistant’s slot becomes available for a fresh setup.
Assistants follow the brand’s type. For a personal brand, where the brand is you as a creator, founder, or expert, the assistant treats the brand as yours. Posts and other writing it starts for you speak in the first person (“I”), including drafts from your voice notes. On a personal brand, the team assistant is for the people who help you run it. For a business brand, writing speaks as the company (“we”) unless you ask for a post in your own name.
The first time you open Assistants, a four-step walkthrough explains the slots, the meter, and the assistant page. It does not create or connect anything.
The channels an assistant can use:
A personal or team assistant has exactly one channel. To move it to a different Slack workspace or Discord server, disconnect the current one under Channel first. Telegram and Microsoft Teams are not offered for new assistants.
New assistants start paused, and connecting the channel turns them on: finishing the Slack or Discord install, or pairing your WhatsApp. There is no separate enable step. Pause is the one off switch: an assistant someone paused stays paused when its channel is reconnected, and one suspended for billing stays suspended until billing is sorted. For Slack and Discord, add channel rules under Channel to choose where it listens.
Assistants created in the dashboard use your browser’s timezone, shown during setup, for their recurring tasks and suggested posting times.
A workspace owner or admin creates the assistant and connects a team assistant’s Slack or Discord. Each member then links their own Slack or Discord account so the assistant knows who is asking. A personal assistant’s owner pairs their own WhatsApp phone. See Link your chat account.
What an assistant does on its own, meaning its settings, channels, focus areas, recurring tasks, and page watches, can be changed by its owner or a workspace admin for a personal assistant, and by workspace admins for a team assistant, the same in chat and on the dashboard. Other members can still ask it for one-off work, add notes, and act on its ideas, drafts, and approvals.
A Slack workspace or Discord server can connect to one Daoco workspace at a time. To move it, use Disconnect on the existing connection on the assistant’s page, then connect it in the new workspace. Disconnecting releases the account, revokes its stored credentials, removes its channel rules and listening buffer, pauses channel digests, and cancels queued messages. Reconnecting starts with a fresh default rule. Pausing an assistant keeps its connections. If an old connection belongs to a workspace that no longer exists, reconnecting clears that leftover connection automatically.
The assistant lives in its channel, so you ask it for things there. Its page in Daoco is for setup and oversight:
- Until its channel is connected, the page shows only the setup steps: connect the channel, and optionally pick the focus areas it keeps an eye on.
- After that, the page has three parts. Needs you lists drafts waiting for your okay, brand threads waiting on a decision, messages it held back, and ideas it saved. Each row opens the draft’s review page, the thread, or the item’s details. When nothing is waiting, it says so in one line. Working on lists the brand threads it started, newest first, with each finished thread’s result. All threads opens the full list. On its own lists everything it does without being asked, one row each with a switch.
- Settings holds Identity, Instructions, Messages, Channel, Members (team assistants only), Memory, and Usage.
What an assistant does
An assistant can answer from allowed brand context, read connected data, remember useful details, and manage its own recurring tasks. A request such as “draft three launch posts for Acme” opens a normal Acme brand thread, monitors the run, and returns the result to the channel. In that thread, the request shows a via Slack, via Discord, or via WhatsApp label, so you can tell it came from the chat app and was not typed in Daoco. Ask how a job is going and the assistant replies with the thread’s status, its plan, the specialists working now, and the latest update. Follow-ups such as “make the second one shorter” stay with the thread that holds the drafts. A separate request starts another thread instead of waiting behind unrelated work. Questions about what a brand already has are answered in the channel, without starting a thread. Ask for a catch-up on a brand and the assistant reports campaigns, drafts waiting, what is scheduled next, what published recently, active projects, connections, and anything waiting on you. It can also read one campaign in detail with its posts, rollout actions, notes, and engagement; list posts by state, campaign, platform, or date; show a draft with its per-platform copy and its images in the conversation; report how published posts performed; read brand documents and articles; summarize the asset library; list available skills; and report website standing across SEO audits, AI-answer visibility, tracked competitors, keywords, and backlink domains. It can also find any of the brand’s threads, including ones started from the dashboard, and read what was asked and answered there, so “what did last week’s competitor research find?” gets an answer in the channel with a link to the thread. These reads never change anything: queueing, scheduling, and publishing still go through the normal approval flow, and changing a draft continues the thread that produced it. The assistant also follows reviews. Ask “is the launch post signed off?” or “who are we waiting on?” and it answers with where the review stands and its link. Ask it to “remind Sam about the launch deck” and it emails everyone that review is waiting on, at most once an hour per person. It never approves or decides for anyone. Starting a review, adding reviewers, or sending a new version is work it hands to a brand thread. Ask what is going out on a given day and the assistant reads that day’s calendar back in the channel: each delivery’s time, platform and account, the post’s first line, its status, and the campaign it belongs to, with a link to the day on the calendar. Say “tomorrow”, a weekday, or a date. Times use the workspace timezone, which the answer names. It can also change when planned posts publish, from the same conversation. “Move the 10 AM LinkedIn post to 2 PM”, “push everything on Thursday to Friday”, or “swap the morning and afternoon posts” retimes those deliveries in place; moving a whole day keeps each post’s local time on the new day, exactly like dragging it on the calendar. When the request is vaguer than that, the assistant shows what it would do and waits for you to agree. Timing is all it changes: it will not write or edit copy, unschedule, approve, or publish, it refuses times that have already passed and posts that already went out, and to undo a move you ask for the earlier time back. The one exception is a draft card, which has its own Approve, Schedule, and Unschedule. Changes appear on the calendar immediately. Anyone in the channel can ask for the day’s schedule or a preview of a move; applying a move needs a linked account. The assistant distinguishes copy you want returned in the conversation from work you want saved in daoco. Ask it to “draft a reply” when you want text back in the channel. Say “save this as a LinkedIn draft” when you want a durable post you can review and publish later. Internal run warnings stay in daoco instead of appearing as a second answer beside the finished copy. After useful work finishes, the assistant may offer one or two concrete follow-ups. For example, three launch scripts can become a campaign with three saved drafts. Nothing is created until you accept. A button click, an unambiguous “yes,” or a numbered reply continues the same brand thread, so the assistant keeps the drafts, assets, evidence, and feedback that produced the result. Finished work includes one direct daoco link when there is something useful to review. Campaigns open in their workspace, a single draft opens in its editor, and asset batches open in the library. The complete answer still appears in the channel. Work can also go the other way. In a brand chat, ask to be sent something — “send this to my assistant”, “text me the final captions on WhatsApp”, “have my assistant ping me when this is done” — and the assistant delivers it to you on the channel you already use it from, with a link back to the thread. Name a channel and it is used when you have a conversation there. Ask to be told when the work finishes and the assistant reports the outcome to the same conversation once the thread is done. Each request sends one handoff to that conversation, even if the runtime retries. The thread stays exactly where it is in the brand sidebar and also appears in the assistant’s thread list. Its status keeps updating even when you do not request a completion message. If you have never messaged the assistant on any channel, or WhatsApp’s 24-hour reply window has closed, the chat says so instead of sending silently. Slack, Discord, and WhatsApp accept supported files and images as input. The assistant can also return images and files produced by the brand thread. Provider limits still apply to file size, format, and message length. For a personal brand, send a bio, portfolio, or other document about you and the assistant adds it to the brand’s shared files. It is read like a file added in Brand Hub, and the brand documents it informs get proposals to review. Tell the assistant what you’re after (“I’m looking for clients now”, “speaking gigs too”) and it updates the same goals Brand Hub shows. Both need a linked account. Voice notes and audio clips are transcribed on arrival and handled exactly like a typed message, so you can send a WhatsApp voice note or a Slack audio clip instead of typing. If a clip cannot be transcribed, the assistant says so and asks you to type the request rather than guessing at it. The source recording stays with the conversation and thread. It is not attached to the social posts the assistant drafts.Voice notes to drafts
Send your assistant a voice note about something you want to say, and it comes back as post drafts that sound like you. Say “turn this into posts”, “make a LinkedIn post out of this”, or just send the voice note with no other ask. A long typed message works the same way. A voice note that asks a question or gives a task is handled like any other message. The assistant hands your words to a brand thread on the post specialist, word for word, as the source to draft from. The drafts keep your phrasing and cut the filler instead of turning it into generic marketing copy. The thread appears in the brand’s chat like any other, and each draft is a normal saved draft you can open in Daoco. When the drafts are ready, each one comes back as a card: the platform, the first lines, an image when the draft has one, and a Preview draft link. The link opens the post the way each platform will show it, on its own page that fits a phone. Anyone in your workspace can open it, and so can people you invite by email from its Share button: they sign in with that email, see only that post, and get no access to your workspace. You can also ask the assistant for a draft’s link, or to share it with someone’s email (“send the LinkedIn one to sam@client.com”). Only a linked member can ask it to invite someone, and invites send no email: you send the link. Every draft waiting on you is also listed under Needs you on the assistant’s page in Daoco. Select one to open its review page, where you can preview it, schedule it, or post it. In any Daoco chat you can also just ask for it: “post the draft my assistant sent this morning” finds the draft and asks before posting, the same as any other post. A draft for several platforms previews each platform’s copy with its own label. You get one card per draft, up to three, with a link to the rest. Each card has three buttons:- Approve marks the draft approved for review. It does not publish or schedule anything. Tap Undo on the confirmation to put the review state back the way it was.
- Tweak asks what should change. Reply with text or a voice note and the same thread revises the draft, which comes back as a new card. Tweaking an approved draft takes the approval back, since the copy changes.
- Schedule suggests the next good time for that platform, at least an hour out and clear of the brand’s other posts on that platform, using the campaign’s posting times when the draft belongs to one. Tap Confirm to put it on the calendar, or reply with another day and time. Nothing is scheduled until you confirm. Unschedule on the confirmation, or asking for it, moves it back to drafts, as does unscheduling it on the calendar.
Brand voice
Drafts from your assistant are written in the brand’s learned voice, the same one brand chat uses. Each voice note you send for drafts also feeds that voice: daoco keeps your words as evidence next to the brand’s posts and rereads them a few minutes later, so the brand voice picks up how you talk. A brand with no post history yet can learn its voice from voice notes alone. Settings > Identity on the assistant page says whether it has learned the brand voice yet and sums it up. To change it, choose Edit in brand.Channel rules
An assistant’s channel always goes to that assistant and its brand. For a team assistant, choose how it behaves in each Slack or Discord channel with the rules under Channel on the assistant page: it can answer every message, only mentions and direct messages, or listen and learn from a channel (listen mode). To cover another brand, create that brand’s own assistant.Link your chat account
The assistant answers anyone in a connected channel, but it only changes things for workspace members whose chat account is linked to their daoco account. Whatever it starts for you, such as a brand thread, a recurring task, or a note, is attributed to you, and approvals you give from the channel are recorded as yours. To link Slack or Discord: For the managed Discord bot, ask in a server channel and mention the assistant. Its direct messages deliver private account links; they do not accept work requests. Slack supports direct-message requests.- In a channel with the assistant, or a Slack direct message, say link me.
- The assistant sends you a one-time link that only you can see: an ephemeral message in Slack, or a direct message in Discord. On other chat apps, ask in a direct message.
- Open the link while signed in to daoco. Check that it names your chat account, then choose Link my account.
Approve work from a channel
When brand work needs approval, the assistant returns the same approval request shown in the dashboard thread. The request names the action and target. Approve or deny it from the supported channel control, or choose Open in Daoco for its review page. One decision resolves the request everywhere and approved work resumes automatically. Everything that waits on you, whether a post draft, an approval, or an idea, has its own review page, and the assistant’s message about it links there. An approval’s review page shows what will happen, who asked, and when, with Approve and Reject for the people who can approve for that brand. Other members of the workspace see the same page without the buttons. Approval and idea pages are for your workspace only and cannot be shared by email. Once an item is answered anywhere, its page shows the outcome, who answered and when where that is known, and no buttons. Only a linked member with permission for that brand can approve, reject, or close an item from a channel. A button press is checked against the person who pressed it, not the person the request was first sent to. The brand’s approval preferences still apply. A channel message cannot widen the assistant’s brand scope or a connection’s read-only setting.Connect your WhatsApp identity
Daoco answers WhatsApp on a small set of business numbers shared by all workspaces; connecting assigns one to the assistant, and its number shows on the WhatsApp channel once connected. Each phone can be paired with one personal assistant per number so incoming messages reach the right one. A personal assistant pairs only with its owner. Anyone else who tries is pointed to the team assistant on Slack or Discord.- Choose Connect my WhatsApp on the assistant’s WhatsApp channel.
- Enter your complete phone number, including the country code, and choose Send code.
- Enter the six-digit code daoco sends to that phone on WhatsApp.
Start from WhatsApp
Choose Start on WhatsApp in the homepage hero to open a chat with the assistant. This option appears when WhatsApp signup is available. You don’t need an account to start. Text the daoco WhatsApp number and the assistant replies right away with three quick questions, one at a time: your brand’s name, whether the brand is you personally (it’s me) or a business (a business), and where to read about it. Send a website or portfolio, a social profile link, or the platform name and handle, such asinstagram @yourbrand or tiktok @yourbrand. Say skip if there isn’t one. Send a source at any point and it takes one quick look (no account, no charge) while it asks the rest, then tells you something specific it found, with a link to save your brand.
The link opens a short sign-in on your phone, then a confirm page that shows the WhatsApp number being linked, partly hidden. Nothing is linked until you choose Connect my WhatsApp; if it isn’t your number, close the page. What the page asks depends on your account:
- New to daoco. You only confirm. Your workspace and brand are created with the name you gave and what the assistant found, and it starts researching your brand properly, as a person or as a business depending on your answer.
- Already have an account. You also choose where this goes, on the same page. If you belong to more than one workspace, choose one before continuing; none is selected for you. Then pick Use an existing brand or Start a new brand. An existing brand keeps its profile, name, and type; your answers and what the assistant found are only the start of your chat. A new brand takes the name and type you gave. A brand whose personal assistant belongs to someone else can’t be picked, and Start a new brand is unavailable once your plan’s brand limit is reached (the page links to upgrading).
- Confirm your brand. It sends a short summary with looks right and fix something. Type or send a voice note with any correction (“we sell to wedding planners, not cafes”) and it updates the profile the same way the dashboard review does, then shows the summary again. Confirming finishes brand setup; you don’t need the dashboard. For a personal brand, it then asks two quick things, each one skippable: what you’re after (reply with the numbers that fit: a new role, clients or freelance work, speaking and media, growing an audience, brand deals and sponsors, selling your own product, plus the role, field, or niche if you like), and a bio, portfolio, or one-pager about your work (a media kit if you’re a creator). The file is read like one added on the dashboard’s sources step, and brand setup finishes once you’ve answered. If you gave your LinkedIn, daoco reads your work history from it, so there is no resume to send.
- Connect your accounts. It offers the accounts you named when you started, one at a time and in the order you listed them, with a link that goes straight to that platform’s sign-in. Links work once and expire after 15 minutes; reply “connect” for a fresh one or skip to move on. If you named no accounts, it offers Instagram, Facebook, and TikTok. It tells you in the chat when a connection lands.
- Pick a first result. It suggests three things to make first. Tap one and it starts the work, or send a voice note about what’s new this week.
Park work on the notepad
Not everything is worth starting the moment you think of it. Tell the assistant “note: the Q3 deck needs the new retention numbers” or “park this for later” and it saves the note instead of opening a thread. Photos, links, and files sent with a note are saved alongside it. When you are ready, ask it to go through the pad: “handle my notes”, “go through the notepad”, “do everything from this week”. The assistant groups related notes, answers what it can in the conversation, and opens one brand thread per piece of real work. If there is more work than it can start at once, it says which notes are still waiting and picks them up next time. A note stays on the pad until the work it was handed to actually finishes, then closes with the outcome. If that run fails, the note goes back on the pad. The Notepad, under Settings > Memory on the assistant page, shows what is parked, what a thread is already working, and what closed in the last 14 days with a link to the work. You can dismiss a note there or put a closed one back on the pad. To add or change a note, tell the assistant. Notes for a single-brand assistant automatically belong to that brand. To run the pad on a schedule, ask for it: “process my notepad every Friday afternoon.” That task runs in the conversation you asked from and reports there.Review ideas
When a recurring task, a scan, or a recap finds something worth doing that nobody asked for yet, the assistant saves it as an idea instead of only mentioning it. Each idea has a one-line title, why it matters now, the evidence it read (a post’s numbers, a competitor page, a campaign, a document), and the request it would run. Nothing starts until someone accepts. In a channel, tap Accept or Not now on the message, or reply “yes” or “not now”. The message also links to the idea’s review page, where a workspace member can read the evidence and choose Accept or Not now. Accepting opens a brand thread with the idea’s request, and the result comes back to the same conversation. Accepting from a channel needs a Daoco account linked to that channel identity. Dismissing does not. New ideas are listed under Needs you on the assistant’s page. Select one to see why it matters, the evidence with links to its sources, and the request it would run. You can edit the request, choose Start it to open the thread, or Dismiss it. Past ideas lists ideas already started, dismissed, or expired, and you can reopen a dismissed or expired one. An idea follows the same visibility rule as memories: one proposed in a direct message or a private channel may quote what was said there, so the assistant only brings it up in that conversation. In Ideas, workspace admins see every idea. Other members see only ideas proposed where the whole team could see them. If the same evidence turns up again, the assistant updates the existing idea instead of adding a duplicate. A dismissed idea stays dismissed. An idea becomes obsolete when the post it was about is published, when the thread it pointed at finishes, or when 30 days pass without an answer. Obsolete ideas show why.What the assistant remembers and who can see it
The assistant keeps durable memories: preferences, facts, contacts, standing instructions, and short summaries of finished work. Where it learned something decides who it will repeat it to:- Public channel: shared. The assistant can use it in any conversation.
- Private channel or group DM: stays in that channel. It never comes up in another channel or in a DM.
- DM: stays with that person. It only comes up in a DM with them, never in a channel, even when they are the one asking.
Listen mode
If Slack was connected before private-channel listening was available, reconnect it to grant access to private channels and group messages where the app is present. Reconnecting also grants email access so full workspace members can match their Daoco account automatically. Guests and Slack Connect members still link their accounts explicitly. In a Slack or Discord channel, the assistant can learn from the whole conversation, not only the messages that mention it. Under Channel on the assistant page, set a channel’s rule to Listen and learn, or choose it when you add a rule for one channel. Listen mode works on one named channel at a time. It isn’t available for direct messages or WhatsApp. When you turn it on, the assistant posts one short message in the channel saying it will remember what’s said there and how to turn that off. It still replies only when someone mentions it, unless speaking up is on. While listening, the assistant stores the channel’s messages as they arrive and reads them in batches: after about 40 new messages, or once the channel has been quiet for a while. From each batch it keeps decisions, preferences, facts about products and campaigns, people and contacts, and deadlines. It skips small talk, open questions, and anything that looks like a password or payment detail. Hearing the same fact again updates the existing memory instead of adding a copy. Messages are treated as things people said, never as instructions to the assistant. What it learns follows the rules above: a public channel’s memories are shared, and a private channel’s stay in that channel. Messages from bots, including the assistant’s own, are never stored. A message deleted before its batch is read is never learned from, and edits replace the stored text. Stored messages are deleted after seven days. Each channel under Settings > Channel shows its mode, when the assistant last learned from it, and how many memories it learned there. Select the count to see those memories in Memory. To stop, switch the channel back to Mentions and DMs. Listening stops right away and any messages not yet read are deleted. Memories already learned stay until you forget them. Reading batches uses tokens from the workspace balance, like other assistant replies.Speaking up
Speaking up lets the assistant answer a question in a listening channel when nobody mentioned it but it already knows the answer. Someone asks “does anyone know when the spring launch is?”, and the assistant replies in the thread with the date it heard earlier and where that came from. Set it under Settings > Channel > Answer questions nobody asked it. The setting covers every listening channel of that assistant:- Off: the assistant replies only when someone mentions it. This is the default.
- Watch only: the assistant judges each question but posts nothing. Where it could have answered lists each question it judged, what it would have used, and whether it would have replied.
- On: the assistant replies when it is confident the reply would help.
Channel digest
A listening channel can get a regular digest: a short summary of the decisions made, open questions and who asked them, deadlines, and what the assistant learned there. When little happened, the digest says so in one line. Each listening channel has a Daily digest in #channel row under On its own. Switch it on, and select the row to pick the time. The time uses the assistant’s timezone. You can also ask in the channel, for example “send a daily digest here at 9am” or “post a weekly digest here on Mondays”. Only a linked workspace admin can set a digest up from a channel. All three ways set up the same digest: asking again for a digest of the same channel in the same place moves its time instead of adding a second one. A digest covers the last day or the last week of stored messages and what was learned from them. If nothing was said and nothing was learned in that window, the assistant posts nothing. A digest is normally posted in the channel it summarizes. You can ask for a public channel’s digest in another channel or a direct message. A private channel’s digest is only ever posted in that channel. Digests appear under On its own with the assistant’s other recurring tasks. To stop one, switch its row off. A digest is an unprompted message, so it waits for the assistant’s messaging hours like other proactive messages. If the channel stops listening or its routing rule is removed, its digests pause, and On its own shows why. Turn listening back on before resuming them. A digest reads the channel’s stored messages when it runs and doesn’t keep its own copy, so a deleted or expired message never appears in a later digest.Run recurring tasks
Ask the assistant for standing work in plain language, such as “every Monday, summarize last week’s campaign results.” The task appears under On its own on the assistant page. Switch it off to pause it, or select it to see the full request and delete it. If the assistant retries the same task creation while handling your message, it reuses that task rather than creating a duplicate. Recurring assistant tasks start brand work through the same chat runtime as dashboard requests. They use the assistant’s allowed brands and the brand’s approval settings. A task run counts as finished when the brand thread it started finishes, not when it was kicked off, so its status and history reflect the real outcome. A run that loses its thread is marked failed after a few hours rather than staying open. A task can also wait on something happening instead of a clock. Phrase it as “whenever” rather than “every”: “whenever a post is published, send me the link and a three-line summary”, “when a campaign is created, draft a launch checklist”, “ping me when something needs approval”. The assistant can watch a post being published, a brand thread finishing a run, a campaign being created, anything new landing in the review queue, and a focus area milestone. When one of those happens, the run receives the ids, a link to the exact item, and a short summary, then does what you asked and reports back in the channel you set it up from. Event tasks appear in the same Your tasks list, marked with what they watch for and when they last fired, and pause and delete the same way. Each one stays quiet for a cooldown window after it fires, so a batch of posts going out produces one message rather than one per post. An event that arrives while the task is still working on an earlier one is deferred and runs next, and only the newest deferred event is kept.Focus areas
Focus areas are rows under On its own on the assistant page. They decide what it watches and does without being asked. There is one for each marketing area the assistant can watch:
Each row shows what the area sends you and when, or what it Needs before it can run. Areas that are off sit under More it can do. An area that is on waits until you message the assistant in its channel, giving it a place to report. Only the people who can change the assistant can change its focus areas.
Turning an area on starts its tasks in the assistant’s channel. They act for the person who turned it on. Turning it off pauses those tasks and keeps their history. Turning it back on resumes them. A task you switched off inside the area stays off. Select an area’s row to switch its reports one by one, change when it tells you, see the last milestone it found, or connect what it needs.
Your usual engagement is the median interactions of your recent posts on the same platform. The assistant waits until it has at least five measured posts there, and calls out each post once. Milestones come from data daoco already collects when it refreshes post analytics, checks keyword rankings, or runs an AI visibility check. Idea tasks save each idea to Ideas with the posts, comments, pages, or AI answers behind it, and nothing starts until someone accepts.
The weekly reviews do the work themselves, so they use credits like any brand thread. Each one reports back in the channel: what it prepared, what is waiting on your approval, and what it found. Opening a pull request or writing a CMS draft waits for your approval, which you can give from the channel, and nothing is published. Without a connected GitHub repository or CMS, the SEO review lists the fixes with the proposed changes, and the article stays in Articles. The morning agenda and weekly recap say when the next review runs and what it will pick up, or that it is off.
A new personal assistant, including one set up from WhatsApp, starts with Social on, plus SEO and AI visibility when a business brand has a website, with their milestone and idea tasks. On a personal brand it starts with Social and Opportunities instead; SEO and AI visibility stay off until you turn them on. Opportunities is not offered for business brands. A new team assistant starts with the same areas on and a weekly recap across them, without a message for every milestone. The tasks start once the assistant’s channel is connected.
When you connect an account that makes an area available, such as Google Search Console for SEO, the assistant asks once in its channel whether to turn it on. It only asks about areas nobody has turned on or off yet. Choose Accept to turn it on or Not now to leave it off. The question also appears under Ideas. You can also ask in the channel, for example “turn on SEO focus” or “stop watching AI visibility”. From a channel, only a linked member who can change the assistant can switch a focus area.
Everyday tasks. Next to the focus areas are tasks that cover every area or none. Switch one on and it reports in the conversation you last had with the assistant; select the row first to pick another conversation. The person who switches a task on is who it acts for.
Switching one off pauses it and keeps its history. Switching it back on resumes the same task rather than starting a second copy. You can also ask for one by name in any connected channel and it lands on the same switch.
Your own. For anything the list does not cover, ask the assistant: what you want it to do, and when, on a schedule or when something happens. Tasks you asked for are listed under On its own with the rest, on personal and team assistants alike.
Skills. A proactive task can follow a skill from your skill library, and every run loads it before working. Say so when you ask (“every Monday, do the competitor pulse using our pulse report skill”). If your workspace has its own copy of a shipped skill, the task follows your copy; a skill you turned off is not used at all. The competitor pulse task ships with a skill already attached.
When it may message you. Under Settings > Messages, choose Only when asked or Can reach out first. Can reach out first is the default, at any time of day. Choose Edit to set the hours it may message you. The first hours you set use your browser’s time zone. Later edits keep the time zone shown, even if you travel. An unprompted message outside those hours waits and goes out when they open (on WhatsApp, as a notification if the 24-hour reply window closed in the meantime). With Only when asked, the work still runs on schedule, and the messages wait for you instead. Held messages show under Needs you. Select one to see what it says and why it is waiting, then Send now to queue it for delivery or Drop it. Scheduled channel digests count as unprompted and wait too. Replies to something you just said are never held, and neither is speaking up, which answers a message someone just posted.
Watch a page
Ask the assistant to watch a public web page and it checks the page on a schedule and tells you when something happens. For example:- “Watch acme.com/pricing and tell me if it changes.”
- “Let me know when this job post disappears.”
- “Ping me if this page mentions beta.”
- “Alert me if this drops below $49.”
Use slash commands in brand chat
Type/ in the dashboard chat input to open the command menu.
Use a command when the destination is clear. Ask naturally when the request needs several kinds of work and daoco should choose the order. Keep a follow-up in the same thread when it depends on drafts, files, or decisions already there.