ClickUp AI meeting follow-through

Simple Guide

What this is

This is an internal how-to, not a ClickUp product page. Use the Simple Guide to get a pilot running. Use the Detailed Guide for the exact prompts, routing table, agent instructions, Automations, guardrails and sources. Use Flow when you need to show the path from a live meeting to a closed task.

Every prompt block has a Copy button. Fill in the routing table before you turn any agent loose. Do not put passwords, payment details or private finance data into ClickUp prompts, Docs or this playbook.

Simple Guide

What you are building

Today, Fathom records the meeting. A follow-through bot then writes a locked-format summary, creates tasks in the right ClickUp location, posts a recap in the right Chat channel, and chases owners each weekday at 9:00 AM PHT until those tasks are closed. You will replace that chain with ClickUp-native tools.

• ClickUp AI Notetaker joins Zoom, Google Meet, Microsoft Teams or SyncUp meetings and writes a meeting notes Doc.
• ClickUp Brain or a Super Agent rewrites those notes into the locked 14-section summary, then creates or updates tasks in the correct Space, Folder and List.
• A follow-through Super Agent drafts the Chat recap and, after the pilot, can post it as a new channel message.
• A daily chase Super Agent, plus Automations, nudges owners on overdue and due-today work until each task is closed.
• A weekly L10 / scorecard agent and a decision log agent cover meeting prep and a standing record of what was settled.

Simple Guide

Minimum setup

This is the shortest path to a working pilot. An admin or Workspace owner must do steps 1 and 2. You can do the rest once those ClickApps are on and your calendar is connected.

1. Confirm the plan and AI credits. AI Notetaker, Super Agents and Autopilot Agents consume AI Credits on most plans. On new plans, Notetaker uses a fixed 180 credits per notetaker hour, and it is unlimited on Enterprise and Max. Unused monthly credits do not roll over. Check Billing with the founder or an admin before a wide rollout.
2. Turn on the ClickApps. An admin or owner activates the AI Notetaker ClickApp in the App Center (Collaboration or All ClickApps). Super Agents also need the AI ClickApp, and an admin controls who can create or interact with Super Agents.
3. Connect the calendar. Connect your Google Calendar or Outlook Calendar (personal connection only). AI Notetaker can then join meetings that have a Zoom, Teams or Google Meet link. Google Meet and Teams usually need no extra setup. Zoom hosts must allow recording, and may need to approve the Notetaker if Zoom is not connected in ClickUp.
4. Set who the Notetaker is for team meetings. For a shared VBI series, choose Shared Notetaker, or Workspace Notetaker if everyone in the Workspace should see the Doc. Set it on All events for a recurring series. This keeps one set of notes and one usage count. If two people both auto-join with Personal Notetaker, ClickUp creates two Docs and bills two usages.
5. Set auto-join on your primary calendar. In Planner, click the notes icon, then Settings. Under Auto-join, select your primary calendar. Auto-join applies to your Personal Notetaker unless an admin has set a Workspace default Notetaker type. Auto-join will not add a Personal Notetaker if a Shared or Workspace Notetaker is already on the meeting.
6. Confirm the routing pick for the pilot series. The routing table now uses real Workspace locations. For the pilot, get a human confirmation of the live List and recap channel, especially where the table says it needs a pick.
7. Create the Meeting Follow-through Super Agent. In Global Navigation, hover over AI and select Create Agent. Paste the follow-through instructions from the Detailed Guide. Share the pilot meeting notes with that agent (from Planner, from the agent profile, or from AI Notetaker settings under Share meeting notes). Keep it in draft mode: it prepares the locked summary, tasks and recap, and waits for you or the meeting owner to approve before posting.
8. Run one meeting, then review. After the pilot meeting ends, open the notes from Meetings Hub, Planner (View AI Notes) or Docs Hub > Meeting Notes. Check the locked summary, every task location, assignee and due date, and the recap draft. Only then add the daily chase, L10 prep and decision log agents.

Simple Guide

Week 1 pilot

Pick one recurring series, for example one unit's L10, not the whole Workspace. Use Shared Notetaker on All events in that series. After each meeting that week:

• Confirm the meeting notes Doc appeared and that the transcript is in it.
• Run or review the follow-through agent. Compare its 14-section summary with the transcript. Invented owners, dates or quotes must be corrected, and the prompt tightened if needed.
• Confirm each new task landed in the right List, with the named owner and the stated due date, and that no duplicate was created.
• Have a human post the recap as a new Chat message, not as a thread reply, until you trust auto-post.
• The next weekday at 9:00 AM PHT, check whether the chase nudge is correct, and that it stops when the task is closed.

If that week is clean, expand to the next meeting type in the routing table. Do not turn on auto-post or Workspace-wide auto-join until the pilot is signed off.

Simple Guide

Limits to know first

• Only Members and Admins can use AI Notetaker. Guests and limited members cannot.
• AI Notetaker is not supported in breakout rooms. If it is removed mid-call, a replacement only captures notes from the moment it rejoins.
• If a meeting is moved to a later date after it has started, you only get notes for the original meeting. Create a new meeting if you still need notes.
• Language is auto-detected. If several languages are spoken, ClickUp tries to transcribe the whole call in the primary language. Meetings that mix English and Filipino should be checked, especially quotes.
• Standard Automations that key off a date with no time fire at 4:00 AM in the creator's timezone. For the 9:00 AM PHT chase, use an Every... trigger set to 9:00, and confirm the creator's timezone in My Settings, or use a Super Agent trigger and test it.
• Threaded comment replies do not trigger Automations. Agents must never post a recap or chase as a reply inside a thread.
• Sharing a Super Agent with all future meeting notes is documented in product messaging and a ClickUp comment dated 27 October 2026. Whether that share applies to every event in a recurring series is something to test.
• This Workspace may also hold personal lists, payroll lists, and work from other companies. A Workspace-wide agent can leak across those walls unless Knowledge and Tools are limited.
• ClickUp Super Agent examples in official help include "reply to thread." That is not allowed here. Recaps and chases are new messages only.

Detailed Guide

AI Notetaker setup

Official name: ClickUp AI Notetaker. Notes land in ClickUp Docs. You open them from Meetings Hub (Meetings in the Home Sidebar), from Planner by clicking the event and choosing View AI Notes, from the notes icon in Planner, from Docs Hub > Meeting Notes, or by asking Brain.

Activate the ClickApp

1. An admin or owner opens the App Center.
2. From Collaboration, or All ClickApps, find AI Notetaker and turn it on.

Connect a calendar

Connect Google Calendar or Outlook Calendar in ClickUp. The Google Calendar integration is a personal connection; no one else in the Workspace can see your calendar through it. After it is connected, meetings with a Zoom, Teams or Google Meet link appear in Planner and can take a Notetaker.

Choose a Notetaker type

If a Shared or Workspace Notetaker is already set, Personal Notetaker is unavailable until someone selects No Notetaker. Anyone invited can add or remove a Shared or Workspace Notetaker. For a recurring event, Planner asks whether the change applies to This event or All events.

Auto-join

1. In the upper-right of Planner, click the notes icon, then Settings. You can also open Workspace settings and choose AI Notetaker. Your options are on the My settings tab. Admins also see Admin settings.
2. In the Auto-join row, select the primary calendars to join automatically. Only primary calendars can be selected.
3. Admins can set the default Notetaker type used when members turn on auto-scheduling, plus the display name, join message (up to 900 characters), and whether notes include the AI Summary, Recording, Transcript and Chat messages. Admins can lock those defaults.
4. Optional, on My settings: Share meeting notes lets you choose a Super Agent that should receive your notes automatically.

Send a Notetaker to one meeting

1. In Planner or Calendar Glance, click a meeting that has a video link.
2. Click Add AI Notetaker and choose the type.
3. At the meeting time, the Notetaker joins as an attendee and posts its join message.
4. For an ad hoc meeting, paste the meeting link from the Planner notes icon, from your avatar menu (AI Notetaker), or from Meetings Hub > Send AI Notetaker.

What the built-in Doc contains

The Doc is named with the meeting title and date. It includes attendees, a recording (video if the call is one hour or less, audio if it is longer; video is in beta), Overview, Key Takeaways, Next Steps, Key Topics, chat history when available, and a transcript (sometimes on a second Doc page). For Teams, chat history is captured only if the meeting chat is open to anonymous participants. Channel-meeting chat is not captured.

Zoom notes

From the Zoom web Settings, Recording tab, the host should enable Record to computer files, Internal meeting participants, and Auto approve their permission requests. If the host has not connected Zoom in ClickUp, they must click Approve for each Notetaker, and can tick Apply these permissions to all future requests in this meeting. ClickUp uses Zoom's native recording notice to tell participants that recording is on.

Retention

On Enterprise and Max, owners and admins can set Workspace retention (Never expires, or 30, 90, 180 or 360 days). Shortening the period permanently deletes older notes and recordings. This is not available on Core or Business.

Detailed Guide

Locked summary format

This is the format the current follow-through bot already uses. Every official recap must use these 14 sections, in this order, with these rules. Do not add sections. Do not skip a section; if a section has nothing, write "None."

• Every section name is bold, on its own line.
• When a section has more than one item, use bullet points, one item per bullet. Settled uses numbered points.
• Bullets are full plain sentences. Never use "Label: value" fragments.
• Use a short paragraph only for one-item sections, the Bottom line, and the Walkthrough paragraphs.
• Use only the meeting notes Doc, the transcript and the recording. Invent nothing. If something was not stated, say so. Mark anything unverified as "not verified".
• Write in English. Keep quotes in the original language and add an English translation when the original is not English.
• Show times in Asia/Manila (PHT). If the meeting was not in PHT, include both the meeting's local time and the PHT equivalent.
• No emoji. No em dashes. Keep personal details out. This summary is for this company only; do not include work from other companies, personal matters, or private finance.

1. Header. One line with the meeting name, the date, the PHT time, who attended, and the ClickUp meeting notes Doc link.
2. Overview. The meeting's purpose, who was there, and the 3 to 5 things that matter.
3. Bottom line. A short paragraph on what the meeting really was and what it means.
4. Settled. Numbered decisions, each with its owner and a timestamp, plus a quote with a translation when it helps. End with what was not settled.
5. Walkthrough: topics. What was actually discussed, topic by topic, in full detail: names, numbers, tools, what is broken, and whether each thing is a draft or final. Write one paragraph per topic. Replace the heading with the short list of topics.
6. What changed. Bullets, each one old versus new, with its source or timestamp.
7. Your commitments. Bullets, written from the principal's point of view when that person attended. If they did not attend, list only commitments assigned to them, or write "None." Replace [Principal] with the name your team uses, in a private copy of the prompt.
8. Others owe you. Bullets, each with the owner, the item and the date.
9. Still open. Bullets.
10. Flags. Bullets: risks, sensitive points, what was kept out of records, and doubts about the source.
11. Exact words. Bullets: short quotes with the speaker and timestamp, in the original language with a translation.
12. Next touchpoint.
13. Decisions for [Principal]. Numbered, at most 3. Example, fill in the principal's name in your private prompt.
14. Sources. Include the meeting notes Doc link, and mark anything unverified as "not verified".

Detailed Guide

Brain prompts

Use these in two places. First, open the meeting notes Doc, click Brain in the upper-right, and paste the summary prompt, then the task prompt. Second, paste the same text into the Super Agent instructions so the work can run after every shared meeting. During the pilot, run them by hand until the output is trustworthy.

1. Rewrite the notes into the locked summary

You are writing the official meeting follow-through summary for Virtual Business Integrator (VBI).

Use only the meeting notes Doc, the transcript, and the recording. Do not invent facts, owners, dates, numbers, or quotes. If something is missing or was not stated, say that it was not stated. Mark anything unverified as "not verified".

Write one summary, in English, using the 14 sections below in this exact order. Do not add sections. Do not skip a section. If a section has nothing, write "None."

Formatting rules:
- Every section name is bold, on its own line.
- When a section has more than one item, use bullet points, one item per bullet. Settled uses numbered points.
- Bullets are full plain sentences. Never use "Label: value" fragments.
- Use a short paragraph only for one-item sections, the Bottom line, and the Walkthrough paragraphs.
- Keep personal details out. Respect the walls between companies. This summary is for VBI only. Do not include work from other companies, personal matters, or private finance. Keep this summary inside the company this meeting belongs to.
- Do not use emoji. Do not use em dashes.
- Show times in Asia/Manila (PHT). If the meeting was not held in PHT, include the meeting's local time and the PHT equivalent.
- Quotes stay in the original language, with an English translation when the original is not English.

Sections:
1. Header. One line with the meeting name, the date, the PHT time, who attended, and the ClickUp meeting notes Doc link.
2. Overview. The meeting's purpose, who was there, and the 3 to 5 things that matter.
3. Bottom line. A short paragraph on what the meeting really was and what it means.
4. Settled. Numbered decisions, each with its owner and a timestamp, plus a quote with a translation when it helps. End with what was NOT settled.
5. Walkthrough: <topics>. What was actually discussed, topic by topic, in full detail: names, numbers, tools, what is broken, and whether each thing is a draft or final. Write one paragraph per topic. Replace <topics> with a short list of the topics.
6. What changed. Bullets, each one old versus new with its source or timestamp.
7. Your commitments. Bullets. Write these from the principal's point of view when that person attended. If they did not attend, write any commitments assigned to them, or "None."
8. Others owe you. Bullets, each with the owner, the item, and the date.
9. Still open. Bullets.
10. Flags. Bullets: risks, sensitive points, what was kept out of records, and doubts about the source.
11. Exact words. Bullets: short quotes with the speaker and timestamp, in the original language with a translation.
12. Next touchpoint.
13. Decisions for [Principal]. Numbered, at most 3. Example, fill in the principal's name in your private prompt.
14. Sources. Include the meeting notes Doc link, and mark anything unverified as "not verified".

After you draft the summary, wait for a human to approve it before you post it anywhere. During the pilot, leave the approved summary in the meeting notes Doc or in a linked follow-through Doc. Do not auto-post.

2. Turn agreed actions into tasks, without duplicates

Paste this after the summary is approved. Replace the routing table in the prompt with the filled-in table from this playbook.

After the locked summary is complete, extract every action item that was actually agreed in this meeting.

For each action item:
1. Search the Workspace first for an existing open task that is the same work. If you find one, do not create a duplicate. Update the existing task: add a comment that it was confirmed in this meeting, link the meeting notes Doc, and adjust the assignee or due date only when the meeting clearly changed them.
2. If no existing task matches, create one task in the Space, Folder, and List from the routing table for this meeting type. If the meeting type is not in the table, do not guess a location. Hold the item and flag it for Charms.
3. Set the task name as a clear verb-plus-outcome sentence.
4. Assign the owner named in the meeting. If no owner was named, leave the task unassigned and flag it in Still open. Do not guess an assignee.
5. Set the due date that was stated. If no date was stated, leave the due date empty and flag it. Do not invent a date.
6. Set priority only if the meeting stated urgency. Otherwise leave the default.
7. Link the meeting notes Doc. Put the source quote or timestamp in the task description.
8. Do not create tasks for ideas that were only discussed and not agreed. Do not create tasks for work that belongs to another company.

Routing table (example, fill in). If a row still says it needs a human pick or is unmatched, do not guess a location. Hold the item and flag it for Charms.
- [Unit] weekly / L10 | [Unit Space] > [L10 Folder] > [To-Dos List] | Recap: #[unit]-l10-recap | Confirm the live L10 folder if more than one exists
- [Unit] Mini L10 | Use only the series chosen for the pilot | [Unit Space] > [Mini L10 Folder] > [To-Dos List] | Recap: #[unit]-l10-recap | Other series may have unmatched recap channels
- Systems | [Systems Space] > [Systems List] | Tickets: [Ticketing Space] > [Tickets List] | Recap: do not auto-choose until a human picks a private or public channel
- Event track A | [Events Space] > [Event Folder] > [Event List] | Recap: #[event]-recap
- Event track B | Treat a second track as its own meeting type | [Events Space] > [Track Folder] > [Track List] | Recap: #[track]-recap
- [Unit] Marketing | [Marketing Space] > [Marketing Folder] > [Topic List], only when the topic is clear | Recap: #[unit]-marketing-recap
- HR Recruitment | [HR Space] > [HR Folder] > [Recruitment List] | Recap: private HR recruitment channel only
- HR Employee Care | [HR Space] > [HR Folder] > [Employee Care List] | Recap: private employee-care channel only
- HR payroll or sensitive employee matters | [HR Space] > [HR Folder] > [Payroll or employee-relations List] only | Recap: the matching private HR channel | Never a public or general team channel
- Leadership team | [Leadership Space] > [Leadership Folder] > [To-Dos List] | Decision log: [Decision Log List] | Recap: do not auto-pick if more than one similarly named channel exists
- Launch / program | [Launch Space] > [Launch List] | Recap: #[program]-recap
- 1:1 | [1:1 List] only when a list for that pair already exists
- Any other 1:1 or unlisted meeting | Unmatched. Do not guess
Never use a personal space, another company's space, an archived or copied space, or leftover sample channels.
Return a numbered list of tasks created or updated, with the List, assignee, due date, and a link to each task. If you skipped an item, say why.

3. Draft the Chat recap

Draft a ClickUp Chat recap for the channel named in the routing table for this meeting type.

Rules:
- Write it as a new channel message. Never post it as a reply inside a thread.
- If more than one update is going out, number the messages (1, 2, 3).
- Do not use emoji. Do not use em dashes.
- Keep it short: meeting name, date in PHT, link to the locked summary, the settled decisions, and the action items with owners and due dates.
- Tag owners with ClickUp mentions in the form [@Name](#user_mention#ID) only when you have the correct user ID. If you do not have the ID, write the person's name in plain text and flag it so a human can add the mention.
- Invent nothing. If an owner or due date was not stated, say that it is not set.
- During the pilot, do not auto-post. Leave the draft for Charms or the meeting owner to approve, then post. After the pilot is signed off, post only the approved text as a new message in the correct VBI channel.

Detailed Guide

Routing table

This table is a set of examples to fill in. Do not put real Space, Folder, List or channel names, or any ClickUp IDs, on this public page. Copy the table into a private Doc or spreadsheet, then replace every placeholder. "Needs a human pick" means more than one live option may exist in your Workspace. "Unmatched" means no dedicated location has been filled in yet. Until a pick is confirmed, agents must hold the task and flag it for Charms.

Do not route here

Keep this list generic on the public page. In your private copy, name the actual locations that must stay out of team routing:

• Personal spaces and personal task lists, including any list that mixes companies or private life with team work.
• Archived lists, copied or "old" spaces, and unfinished placeholder lists.
• Spaces, lists and channels that belong to other companies.
• Leftover sample or tutorial channels.
• Public or general team channels, when the work is HR, payroll, sensitive employee matters, or internal systems that should stay private.

Routing table (example, fill in). Put real names and IDs only in a private copy.

[Unit] weekly / L10 | [Unit Space] > [L10 Folder] > [To-Dos List] | Recap: #[unit]-l10-recap | Confirm the live L10 folder if more than one exists | Unit decision log may be unmatched

[Unit] Mini L10 | [Unit Space] > [Mini L10 Folder] > [To-Dos List] | Recap: #[unit]-[series]-l10-recap | Pick one series for the week 1 pilot | Recap may be unmatched for other series

Systems | [Systems Space] > [Systems List] | Tickets: [Ticketing Space] > [Tickets List] | Recap: needs a human pick between public and private | Do not use a personal Systems list

Event, track A | [Events Space] > [Event Folder] > [Event List] | Recap: #[event]-recap

Event, track B | [Events Space] > [Track Folder] > [Track List] | Recap: #[track]-recap | Treat as a separate meeting type

[Unit] Marketing | [Marketing Space] > [Marketing Folder] > [Topic List] | Recap: #[unit]-marketing-recap | Default List and duplicate homes need a human pick

HR Recruitment | [HR Space] > [HR Folder] > [Recruitment List] | Private recruitment channel only

HR Employee Care | [HR Space] > [HR Folder] > [Employee Care List] | Private employee-care channel only

HR payroll or sensitive employee matters | [Payroll or employee-relations List] only | Matching private HR channel | Never public or general team channels

Leadership team | [Leadership Space] > [Leadership Folder] > [To-Dos List] | Decision log: [Decision Log List] | Recap: pick the live channel if more than one exists

Launch / program | [Launch Space] > [Launch List] | Recap: #[program]-recap

1:1 or unlisted | [1:1 List] only if one exists for that pair | No general fallback unless you create one | Do not guess

Do not route to personal spaces, other-company spaces, archived or copied spaces, leftover sample channels, or public channels for HR or sensitive work.

Detailed Guide

Agents overview

ClickUp has two agent types you will use. Help articles still call the trigger-based ones Autopilot Agents (formerly Custom Agents). Billing pages may also call them AI Automations. Super Agents are the workspace-wide teammates.

Create four Super Agents first. Use Autopilot Agents and standard Automations to glue reminders and status changes around them. An admin must allow your role to create Super Agents. Share each agent with the people who need to trigger it. If someone must add the agent to meetings in Planner, test whether they need manage access, not only trigger access. That permission detail is unverified in official help and should be tested.

ClickUp also sells Certified / catalog agents, including an Action Item Extractor in the Accelerator package. Those are optional and may require a sales conversation. This playbook assumes you build the four agents below yourself.

Detailed Guide

Meeting Follow-through agent

This is the core replacement for the current post-meeting bot.

• Trigger. A meeting notes Doc is shared with the agent (from Planner, from the agent profile for all existing and future notes, or from AI Notetaker settings > Share meeting notes). You can also @mention the agent on the Doc during the pilot.
• Tools and permissions. Read Docs (meeting notes and the routing table Doc if you store one). Search tasks. Create and update tasks. Post or draft Chat messages. Link Docs to tasks. Do not give it access to other-company Spaces, personal spaces, or any finance, payroll or password Docs.
• Knowledge. This playbook's locked format, the filled routing table, and the VBI Lists and Chat channels in that table only.
• Guardrails. Draft mode in week 1. Never reply inside a thread. Numbered messages. No emoji. No em dashes. Invent nothing. Search for duplicates before creating a task. VBI only.

You are the VBI Meeting Follow-through agent.

When you receive a new VBI meeting notes Doc, do this work in order:
1. Read the Doc, the transcript, and the recording if present. Use only those sources.
2. Write the official 14-section locked summary using the locked summary prompt stored in your instructions. Put the draft in the meeting notes Doc or a linked follow-through Doc.
3. Extract agreed action items only. Search for existing open tasks first. Create or update tasks in the Space, Folder, and List from the routing table. Assign the named owner and the stated due date. Do not invent owners or dates. Do not create duplicates.
4. Draft the Chat recap for the channel in the routing table, following the recap prompt.
5. Append settled decisions to the decision log named in the routing table, or flag that the log location is not filled in.
6. Stop and wait for human approval before posting the recap or sending mentions. During the pilot, never auto-post.

Hard rules:
- Stay inside this company. If the notes belong to another company, mix companies, or belong in a personal space, do not act. Tell Charms instead.
- Never post as the founder or principal, and never post a message that will go out under that person's name, without their explicit approval for that message.
- Never post as a reply inside a thread. Recaps are new channel messages only. ClickUp's own Super Agent examples sometimes reply in a thread; you must not.
- If more than one message is needed, number them.
- No emoji. No em dashes. No secrets, passwords, payroll figures, or private finance data.
- HR, payroll and sensitive employee items stay in the private HR Lists and channels in the routing table. Never post them to public or general team channels.
- If the meeting type is unmatched, or the row says it needs a human pick, do not guess the List or channel.
- Tag people with [@Name](#user_mention#ID) only when you have the correct user ID.
- If you are unsure, ask Charms. Do not invent a decision, owner, or due date.

Detailed Guide

Daily owner chase

This replaces the weekday 9:00 AM PHT follow-through chase. It runs until each meeting-created task is closed. Pair it with Automations so a due-date comment still goes out even if the agent is paused.

• Trigger. Every weekday at 9:00 AM Asia/Manila. In a Chat channel, create a Chat Automation with the Every... trigger, set the time to 9:00, and confirm the timezone belongs to Asia/Manila (the Every... trigger uses the timezone in the creator's My Settings unless the trigger shows its own timezone field). Then launch this Super Agent, or @mention it. Test the first run on a weekday. Do not rely on "Due date arrives" or "Date is Before/After" for 9:00 AM; those fire at 4:00 AM local time when no time is set.
• Tools and permissions. Search tasks in the routing-table Lists only. Read status, assignee, due date and last comment. Draft or post a new Chat message, or a new comment on the task. Do not close, delete, reassign or change due dates on your own.
• Guardrails. Chase only tasks this system created or updated from a meeting, or tasks tagged for follow-through (agree the tag with the team, for example "ft-chase"). Skip closed or done tasks. One nudge per task per day. Never a thread reply. Draft mode in week 1.

You are the VBI daily owner-chase agent.

Every weekday at 9:00 AM Asia/Manila (PHT), review open follow-through tasks in the Lists from the routing table.

Include a task only if:
- it is open (not closed or done), and
- it was created or updated from a meeting follow-through, or it has the agreed follow-through tag, and
- it is overdue, or it is due today in PHT.

For each task, draft one short nudge:
- Address the assignee by name. Use [@Name](#user_mention#ID) only when you have the correct user ID.
- State the task name, due date in PHT, List, and link.
- State whether it is due today or overdue, and by how many days if overdue.
- Ask for a status, a new date, or a close.
- Do not use emoji or em dashes.
- Do not invent progress. If there is no recent comment, say that no update was found.

Posting rules:
- Group nudges into a single numbered Chat post in the agreed follow-through channel, plus a comment on the task if that is enabled after the pilot.
- Always post as a new message or a new task comment. Never reply inside a thread.
- If more than one message is required, number them.
- During the pilot, leave a single draft for Charms to approve. Do not auto-post.
- If there is nothing to chase, write "No follow-through tasks due today or overdue" and do not ping anyone.
- Do not close, delete, reassign, or change a due date. If an owner asks for a change, flag it for Charms or the meeting owner.

Detailed Guide

Weekly scorecard / L10 prep

Use this the day before each L10 or weekly scorecard meeting, once the routing table has the right List and channel.

• Trigger. Every... on the agreed weekday and time (set in PHT), or a manual @mention the day before the meeting. Adjust the day to match each series. Test the first run.
• Tools and permissions. Read the unit's scorecard, rocks, open issues, overdue follow-through tasks, last meeting's locked summary and the decision log. Draft a prep Doc and a short Chat note. Do not change scorecard numbers.
• Guardrails. Draft only until signed off. Invent nothing. If a number is missing, say it is not in the source. VBI unit only.

You are the VBI weekly L10 and scorecard prep agent.

When triggered for a named unit (example, fill in: a marketing team, an HR team, a systems team, or a program team), prepare a pre-read for the next L10 or weekly scorecard meeting.

Use only: the unit's scorecard, rocks, open issues list, open follow-through tasks, the last locked meeting summary, and the decision log. If a source is missing, say it is missing. Do not invent metrics, owners, or dates.

Write a prep Doc with these headings, in this order:
1. Header. Unit, meeting date and time in PHT, and links to the sources you used.
2. Scorecard. What is on track, what is off track, and what was not found in the source.
3. Rocks. Status of each rock as stated in the source.
4. Open issues. Issues still open, with owners if stated.
5. Carry-over actions. Open or overdue follow-through tasks, with owners and due dates.
6. Decisions still open. Items from the last summary or decision log that were not settled.
7. Suggested issue list for this meeting. Only items supported by the sources.
8. Sources. Links, and anything that is not verified.

Also draft a short Chat note pointing to the prep Doc. Do not use emoji or em dashes. Do not post as a thread reply. During the pilot, leave both drafts for Charms or the meeting owner to approve. Do not change scorecard numbers or close issues on your own.

Detailed Guide

Decision log agent

Keep one decision log per VBI unit, as a Doc or a List, and put that location in the routing table. The follow-through agent can append after each meeting. This agent is the specialist if you want a second pass or a weekly cleanup.

• Trigger. After a locked summary is approved, or a weekly Every... review.
• Tools and permissions. Read the locked summary. Append a row or bullet to the named decision log. Do not overwrite earlier entries.
• Guardrails. Log only items in Settled. Include the date, owner, meeting link and what was not settled. If the log location is empty, stop and tell Charms.

You are the VBI decision log agent.

When a locked meeting summary is approved, append that meeting's settled decisions to the decision log named in the routing table for the meeting type.

For each settled decision, record:
- Date and time in PHT
- Meeting name and link to the notes Doc
- The decision, in a full sentence
- The owner
- The source timestamp or quote, if present
- Anything the summary listed as not settled

Rules:
- Append only. Never edit or delete earlier entries.
- Log only what appears in Settled. Do not promote Still open items into decisions.
- If the routing table has no decision log location, do not create a new Doc or List. Flag it for Charms.
- No emoji. No em dashes. No secrets or private finance data.
- VBI only. If the meeting is not a VBI meeting, do not log it here.
- During the pilot, draft the new entries and wait for approval before writing them into the log.

Detailed Guide

Automations

Standard Automations sit on a Space, Folder, Subfolder or List. Chat Automations sit on a Chat channel (robot icon > Create Automation). Date-based Automations use the timezone of the person who created them. If the date has no time, they fire at 4:00 AM local time.

Plan limits for standard Automation actions, from ClickUp's public pricing page: Free, 100 actions per Workspace per month and 5 active rules; Core, 1,000 actions and unlimited rules; Business, 10,000 actions; Enterprise and Max, 1 million actions. Advanced Automations (webhooks, email, SMS, integrations) start at Business.

Useful actions, from ClickUp's Automation Actions help: Add a comment, Send a channel message, Change status, and for Autopilot Agents: Launch Autopilot Agent, Create Docs with AI, Create task with AI, Create subtask with AI, and Edit Custom Field with AI. Confirm the exact action names in your Workspace when you build the rule, because menus can vary by plan.

Detailed Guide

Digests and movement reviews

The current system also produces integrator-style daily movement reviews, a daily digest, and a decision log. The decision log agent covers the third. Use this prompt for a short weekday digest in the agreed VBI channel. Keep it in draft mode until you like the tone.

You are writing a short VBI daily movement digest for [UNIT or all routed VBI units] for [DATE] in Asia/Manila (PHT).

Use only ClickUp tasks, comments, Chat posts, and meeting follow-through Docs from the last 24 hours in the routing-table locations. Do not invent work. If nothing moved, say so.

Write one numbered Chat message with these headings, in this order:
1. Header. Unit, date in PHT, and the time window covered.
2. Closed since yesterday.
3. Still open and due today or overdue, with owners.
4. New follow-through tasks created from meetings, with Lists and owners.
5. Decisions logged in the last 24 hours.
6. Flags. Blockers, missing owners or due dates, and anything not verified.
7. Sources. Links only.

Rules:
- Full sentences. No "Label: value" fragments.
- No emoji. No em dashes.
- Post as a new channel message, never as a thread reply.
- VBI only. Do not include other companies or personal finance.
- During the pilot, leave a draft for Charms to approve.

Detailed Guide

Guardrails

Put these rules in every agent. They match how VBI already runs follow-through in ClickUp Chat.

• Draft first. Auto-post only after the pilot is signed off.
• Never reply inside a thread, including a recap thread. New message or new task comment only.
• Number messages when more than one goes out (1, 2, 3).
• No emoji. No em dashes.
• Invent nothing. Mark unverified items as "not verified". If an owner or date was not stated, leave it empty and flag it.
• Search for an existing open task before creating one.
• Stay inside this company. Do not write into personal spaces or other-company Spaces, and do not mix companies in one recap.
• Keep personal details, passwords, and private finance data out of Docs, tasks, prompts and Chat.
• Use ClickUp mentions only when the user ID is known: [@Name](#user_mention#ID).
• Do not close, delete, archive, reassign or change due dates unless a human asked for that change.
• If two Personal Notetakers would join the same meeting, switch to Shared Notetaker so you do not double usage. Do not use Workspace Notetaker for HR, payroll or sensitive employee meetings.
• Never post under the founder's or principal's name, or as that person, without their approval of that exact message.
• HR, payroll and sensitive employee items stay in the private HR locations in the routing table. Never post them to public or general team channels.
• Do not search or write into personal spaces, archived or copied spaces, or locations that belong to other companies.
• If a route is unmatched or needs a human pick, hold the work and flag Charms. Do not guess.

Detailed Guide

Privacy and walls

A Workspace may also hold personal lists, payroll lists, and work from other companies. A Super Agent with Workspace-wide knowledge could leak work across those walls. Limit each agent's Tools and Knowledge to the Spaces and channels in your private routing table.

• HR, payroll and sensitive employee work stays in the private HR Lists and channels you fill in. Do not put those items in public channels, in a general team channel, or in a mixed meeting recap.
• Do not use Workspace Notetaker for HR, payroll or sensitive employee meetings. Use a Shared Notetaker limited to the invited attendees, or a Personal Notetaker for the HR owner, and share the Doc only with people who already have access to that HR List.
• Never post under the founder's or principal's name, and never send a Chat message or comment that will appear as that person, unless they approved that exact text. In week 1, a human posts the recap. If a Super Agent posts as the person who built it, test that identity before any auto-post. If the builder is the founder or principal, keep the agent in draft mode until a different posting identity is arranged.
• The ClickUp sweep that answers messages to the founder or principal is not in the current four agents. If you build it later, it must draft only, number the drafts, and wait for that person to approve before anything is posted in their name.

Detailed Guide

Knowledge and memory

Give each Super Agent a small, explicit knowledge set. Do not turn on "all Workspace Docs." Attach at least:

• This playbook, or a ClickUp Doc copy of the locked format, guardrails and routing table.
• Your private routing table, including List and channel names, unmatched rows, and the do-not-route list.
• The company decision log once a human confirms it is the live log. The agent should read it before creating tasks, so it does not reopen a settled decision, and should append only after approval.
• The unit's L10 notes Doc or L10 template Doc, when you are prepping or following through on that series.
• The live scorecard, rocks and issues Lists for that unit only.

If the Workspace has an operating-system or wiki Space, and also a copy of that Space, do not attach either one until a human says which is live. Do not attach payroll Docs, employee-relations Docs, personal spaces, or Docs that belong to other companies.

If you enable Super Agent Memory, keep it on for style and repeated routing corrections only. Do not let memory become a second source of facts. Facts come from the meeting Doc, the transcript, the routing table and the attached knowledge Docs. Review memory during the weekly pilot check so it has not stored a wrong List or a detail from another company.

Detailed Guide

Grok to ClickUp map

The goal is to run this work in ClickUp with AI Notetaker, Brain and Super Agents, instead of an external recorder plus outside AI bots. This table shows what already has a ClickUp design in this playbook, and what does not.

Detailed Guide

Owners, KPIs, training

Who owns each agent

If an agent fails (no notes Doc, tasks in the wrong List, a missed chase, or a credit pause), Charms is the first owner to notice and fix it. Log the failure in a place the founder or the systems owner names, such as an existing [Tickets List] or a dedicated failure List.

Pilot success measures

Use these on the single pilot series for two weeks, then review with the founder before expanding. Targets are starting points, not billed promises.

• Every pilot meeting has a notes Doc, with a transcript, within one hour of the end of the call.
• Every agreed action lands in the correct List, or is flagged instead of guessed. Target: no guessed locations.
• Zero duplicate tasks, and zero invented owners, due dates or quotes.
• Every recap is approved by a human before it is posted, is a new channel message, and is never a thread reply.
• Zero posts under the founder's or principal's name without their approval of that text.
• Zero HR, payroll or sensitive employee items in a public or non-HR channel.
• The weekday chase runs at 9:00 AM PHT, and it stops when the task is closed.
• Credit use for the pilot series stays inside the budget the founder sets, and is read from the Super Agent cost panel each week.

Training and SOP for the team

• Before the first pilot meeting, Charms tells the series attendees, in the recap channel, that ClickUp AI Notetaker will join, that they should not remove it, and where the notes will land.
• Meeting owners get a short SOP: how to open the Doc, how to approve or correct the locked summary, how to approve the recap, and that they must not ask the agent to post in a thread or under the founder's name.
• HR meetings use the HR Notetaker and List rules above. Other teams are told not to discuss sensitive employee or payroll matters in a general meeting.
• Do not roll out Workspace-wide auto-join until the pilot SOP has been used on one series and the success measures above have been reviewed.

Detailed Guide

Gaps and next steps

These are the gaps that would stop "the company runs entirely in ClickUp with AI employees" if they are left open. They are ordered so the first ones should be decided before a wide rollout.

1. The Workspace may have more than one live-looking home for the same unit. There may be older and newer L10 folders, several Mini L10 series, and similarly named recap channels. Until those picks are made, auto-create will put work in the wrong place. Next step: the founder or the unit owner confirms the live List and channel for the pilot series. Owner: the founder, with Charms preparing the choices.
2. HR and company walls are not yet encoded as agent permissions, only as prompt rules. Prompt rules are not enough if the Super Agent can search the whole Workspace, including personal lists, payroll lists and other-company lists. Next step: lock Tools and Knowledge to the approved locations, keep HR agents separate, and test that a general agent cannot see HR or other-company Docs. Owner: Charms to configure, a Workspace admin to confirm permissions, the founder to approve the access model.
3. Nothing may post under the founder's or principal's name without their approval, and that identity is untested. Draft mode is in the prompts, but we have not confirmed whether a Super Agent posts as itself, as Charms, or as the connected user. Next step: test posting identity on a private dummy channel, and keep approval required for any message in the founder's name. Owner: Charms to test, the founder to sign off.
4. Two current bot jobs have no ClickUp agent yet: the Integrator daily movement review, and the ClickUp sweep that answers messages to the founder or principal. The digest prompt is not a full Integrator review. The sweep is a gap. Next step: the founder decides whether movement review stays ClickUp-only or still uses Slack, WhatsApp and email, and whether a draft-only sweep agent should be built. Owner: the founder. Charms should not build the sweep until that decision is made.
5. There is no Task Manager agent. Meeting follow-through will not keep the wider backlog clean. Next step: decide if this is in scope for the ClickUp AI-employee set, and if so, which Lists it may touch. Owner: the founder, with Charms proposing a later prompt if the answer is yes.
6. Unit decision logs, a follow-through tag, a failure List, and a general 1:1 fallback List may be unmatched or missing. Next step: confirm the company decision log, then decide whether unit logs, a chase tag, a failure List and a fallback 1:1 List should be created. Owner: Charms proposes, unit owners and the founder confirm.
7. Cost is not yet budgeted against the real plan. Official help prices extra AI Credits at $0.001 each, in $10 increments, and estimates 100 to 300 credits per Super Agent run, plus 180 credits per Notetaker hour on metered plans. Super Agents still spend credits on the highest plan. Next step: an admin reads the current plan and credit balance in Billing, then the founder sets a pilot budget. Owner: admin to read Billing, the founder to set the budget. Seat dollar prices remain unverified here.
8. Sharing meeting notes with a Super Agent on a recurring series, and the 9:00 AM PHT chase trigger, are still tests, not proven in this Workspace. Next step: run them on the pilot series in week 1 and record the result in a private copy of this playbook. Owner: Charms.

Detailed Guide

Rollout checklist

Status below is the planned sequence as of 12 November 2026, Asia/Manila. Update it as you complete each step. The boxes show recorded status, not a live form.

Before the pilot

• Admin confirms plan, AI Credits, and whether the Workspace is on the new plans or still on a Brain / Everything AI add-on.
• Admin turns on the AI Notetaker ClickApp and the AI ClickApp, and allows Charms to create or manage Super Agents.
• Calendar connected. Zoom host settings checked if the pilot series is on Zoom.
• Routing table filled for the pilot meeting type, including the recap channel and decision log location.
• Shared Notetaker (or the agreed type) set on All events in the pilot series.
• Follow-through Super Agent created, notes shared with it, draft mode on.

Week 1

• Run the first meeting. Confirm the Doc, transcript and recording (if enabled) landed in Meetings Hub.
• Review the locked summary against the transcript. Fix invented or missing items and tighten the prompt if needed.
• Confirm task routing, assignees, due dates and no duplicates.
• Human posts the recap as a new Chat message, with numbered mentions where IDs are known.
• Next weekday, 9:00 AM PHT: test the chase draft, then the Automation timezone.
• Append a draft decision log entry and review it.

After a clean week

• Turn on auto-post for recaps only if the founder or the meeting owner signs off.
• Add the daily chase, L10 prep and decision log agents, still in draft for their first run.
• Expand to the next meeting type in the routing table, one series at a time.
• Only then consider Workspace-wide auto-join or a Workspace default Notetaker type.
• Watch AI Credit use after the first two weeks, especially Notetaker hours at 180 credits each.

Detailed Guide

Verified vs test

Detailed Guide

Plan and credits

Workspaces are being moved to ClickUp's newer plans (Free, Core, Business, Business Plus, Enterprise, Max). Some Workspaces still use older plans and may still have a Brain AI or Everything AI add-on. Check which one VBI is on in Billing before you promise a rollout size.

AI Credits on the new plans

Credits are pooled across the Workspace, reset on the first of each month, and do not roll over. When they run out, credit-consuming features pause until you buy more or upgrade. Brain chat and credit use are tracked separately, but starting a credit feature from a Brain conversation still spends credits. Every Super Agent run spends credits, even when it only answers a question.

If a Workspace on the new plans still has an older AI Notetaker add-on, usage may be measured in hours instead of credits. Usage itself is measured in minutes. Confirm in Billing.

Older plans and add-ons

If VBI has not been upgraded: paid plans with no AI add-on get 1,000 credits per user that do not reset. The Brain AI add-on includes 1,500 credits per user per month and does not include AI Notetaker. The Everything AI add-on includes 5,000 credits per user per month and includes AI Notetaker. An AI Notetaker add-on, where it still exists, has been described as 20 hours to start, then a purchased block of hours. Dollar prices for those add-ons are unverified here; read them in Billing.

What to watch

• A one-hour Notetaker session costs 180 credits on metered plans. Two Personal Notetakers in the same meeting cost twice that.
• Super Agents never become unlimited, including on Max.
• Seat prices from the public pricing page were not reliably readable in this pass, so they are not quoted. Check Billing with the founder or an admin.
• Official help, checked again on 12 November 2026: extra AI Credits are $0.001 per credit, sold in $10 increments (10,000 credits). A Super Agent or Autopilot run is estimated at 100 to 300 credits, more if it takes more actions. Every run spends credits, including answers with no action. Admins and the agent's managers can hover Credits on the Super Agent profile to see cost and usage.
• A rough pilot sketch, not a quote: one metered Notetaker hour is 180 credits; one follow-through run may be 100 to 300; a weekday chase and digest can add another 100 to 300 each. Ten meetings in a week, plus daily chase and digest, can use several thousand credits before L10 prep. Set a budget after Billing is read. This sketch is unverified against VBI's actual plan and meeting volume.

Detailed Guide

Sources

Official ClickUp pages opened on 12 November 2026, Asia/Manila. Open a link if you need to re-check a limit or a button name, because ClickUp changes these menus.

Flow

End-to-end flow

This is the path from a live VBI meeting to a closed task. During week 1, the human approval gates stay on. After the pilot is signed off, the same path can post the recap and the weekday chase without a draft step.

Automatic once set up   Human approval in the pilot

A second, slower loop sits beside this path: the weekly L10 / scorecard prep agent reads the last summary, open tasks, scorecard and decision log, and drafts a pre-read the day before the next meeting.

Flow

Who does what

Questions about this playbook go to the founder or the meeting owner. Do not put credentials or finance files into the agents, the routing table or this page.