Header Logo
(802) 372-1377
info@tuckconsultinggroup.com
Tuck Consulting Group
Book A Call

How to Manage All Your Work in ClickUp with AI Agent

Summary: ClickUp AI transforms how teams manage work by automating repetitive processes, streamlining communication, and turning unstructured data (like those pesky emails) into actionable tasks. In this article, we’ll explore how to leverage ClickUp’s AI Agent to route and manage emails directly within ClickUp, creating a centralized, intelligent workspace that keeps your team organized and focused.

Turn Emails Into Auto-Assigned Action Items and Manage Your Inbox with ClickUp AI

 

At Tuck, our nonprofit and small business clients routinely share a big problem: they’re missing a single source of truth for their operations or their client tasks, a central location where teams put their “work”. With a premier work management tool like ClickUp, you can reduce task switching and foster true team collaboration by making all of your work — support tickets from a website form, internal requests via Slack or Teams, or client emails in Outlook — visible in one place.

 

Many of our clients ask us to help take that first step with a question like this: “How do we stop using email forever?” That may be a stretch goal, but it’s not difficult to route emails into specific folders in ClickUp, plan follow-up actions, and situate that work among the other priorities they already manage there. And best of all, this same approach can apply to other data sources like deals in HubSpot or Slack messages, and even help supercharge ClickUp’s native integrations.

 

Our approach starts with ClickUp’s Email to List feature to forward emails directly into a ClickUp list. Then, we use AI Fields to read and assess the messages, and level it up with AI Agents to create automatically assigned follow-up tasks. With these features, we can manage our emails right from within ClickUp, where it’s simple to delegate and track action items, so nothing falls off the radar. In this article, you’ll see exactly how we set this up, where AI adds value, and the best practices and tricks to tailor it for your organization.

 

How to Turn Emails Into Action Items with ClickUp: 

  • In the beginning, your customers created email. When those emails contain work requests or other desired actions, we need a process to turn words into tasks, assigned to the right people for the job. With the setup below, your emails will create tasks in ClickUp. Those tasks will create summaries and lists of action Items, and the action items create auto-assigned sub-tasks for the people to do their work.

    At a high-level, the setup works like this:
    • An email is automatically forwarded (based on customizable criteria) from a specific inbox to a specific ClickUp List.
    • ClickUp automatically creates a task in the list with the email subject as the task name, the email message body in the task description, and any attachments uploaded.
    • ClickUp AI Fields and automations read the task description, and identify specific pieces of information to copy into various fields/columns, such as a message Summary, suggested Action Items, and the Client Name.
    • Then a ClickUp AI Agent reads the summaries and action items, automatically creates sub-tasks for them, and assigns each to the appropriate team member.
      • The agent continues to update action items and subtasks as comments or other updates are made to the task. If a user comments they completed an item, the agent can even close the task automatically.

     

    This enables us to triage email requests alongside our more proactive work in ClickUp, delegate tasks immediately, and see all the component work completed in on place. Now, here’s how we approached building this system.

Want Our Best Tips for Using ClickUp AI?

Instead of rehashing our top tricks, read our blog ClickUp AI Power User Hacks to learn why we always:
  • Start Small and Iterate
  • Include a Design Template or Reference
  • Use Dynamic Fields in Prompts
  • Let AI Help Write the Instructions (Metaprompting)
  • Create an Initialization Prompt
  • Undercut AI Overconfidence

 

Now let’s get to building.

How to Use ClickUp AI Agent to Automate Action Items From Emails

1) Create the ClickUp List and Setup Email Forwarding Rules

 Before we do anything in our email client, we need a destination list first, so we can reference the specific email address ClickUp assigns it and forward messages there.

 Begin by creating a list in the space/folder that best suits the types of messages you’re managing (e.g. Client Emails). Once created, right-click the list in the ClickUp sidebar and select Email to List from the menu to copy its unique address:

create-tasks-by-email

Caption: ClickUp’s ‘Create tasks by email’ feature generates a unique address for each list. Forwarding emails to this address automatically creates new tasks, centralizing requests from any email platform.

For now, we’ll work with one destination list. If you’re perhaps working with different clients or product teams, you can create separate ClickUp Lists for each and configure the forwarding rules in to determine which messages go where. You could also use ClickUp automations or Agents to move tasks into other areas after they’ve been reviewed.

 In any case, with the email address in hand, we can setup auto-forwarding rules in your email client to send any relevant messages into this list.

 For Outlook users, a simple entry point is to right-click an email:

  • Select Rules > Create Rule
  • Set the destination email address and filter criteria
In Gmail:
  • Click the Preferences icon on the right side of the top Search bar, click “Create Filter” to set criteria
  • Choose what happens next, including forwarding the email to ClickUp
How do I set up auto-forwarding rules?
Read more information on how to set up auto-forwarding in Outlook or in Gmail.

 

Can I set up ClickUp Tasks from messages in Microsoft Teams or Slack?
Similar functionality exists to create ClickUp tasks from messages in Teams or Slack, or many other tools, which can be managed just as outlined below.

2) Identify Outcomes, Setup AI Fields and Prompts

Now, we decide what information we want to see and what we want to happen next (e.g. what fields to extract, what to ignore, and how to handle quirks like signatures or multiple addresses).

 

Two initial reminders: 
  • The email message is contained in the task description in plain text.
  • The subject line automatically becomes the task name. Though you can now edit the task names via automations!
  • Identifying basic email fields (sender, send date, CC) is simple. Each field is labeled in the description and easy for AI to read.
FW for potential engagement screenshot- graphic

The beauty of this approach is flexibility. Even more qualitative information, like the message sentiment or a desired follow-up, can be extracted using natural language in the prompt.

 

In our case, these are the fields we identify and extract via ClickUp AI Fields:
  • Email Subject line (this becomes the task name)
  • Sender’s name
  • A summary of the main message
  • Key action items called for by the message
  • Suggested Assignee(s) for each action item
  • Estimated minutes to read the message (the effort to review)
  • Estimated hours to complete the action items (the effort to resolve)

3) And finally, we use an AI Agent for the last step:

  • Auto-generating and assigning subtasks for each action item.
Once you’ve identified the data points you want to report, determine the AI Field options you need.

 

In addition to naming the field and providing prompt text, each AI field has these options to specify:
  1. Template Type (of the prompt): Custom Instructions, a Summary, Progress Updates, Translation, or Action Items.
    1. Custom Instructions allow us to tailor the prompt to our goals. The other options are pre-built prompts to create the generic outputs stated, no need to leverage them here.
  2. Format (of the output text): Bulleted list, short text (i.e. brief), or traditional text (i.e. less brief).
  3. Automations (when to fill/update the field): Autofill when tasks are created (one time) and/or Auto update when tasks change (any time).
    1. You can check both options to ensure fields are set upfront and as the task is changed. Just be sure not to check “Auto update” if you want the field to remain static once it’s generated.
Note also that only text fields (short or long) and dropdown fields support AI in this way. So for the Estimated Minutes and Hours fields above that would traditionally be Number fields (constrained to numerals), we utilized text fields and prescribed the output to be only the numerals.

 

Here are the prompts we utilized for the Email Summary and Action Items fields:

Prompt Examples

1. AI Field Prompt: Email Message Summary (Custom instructions, Long text, Auto-Fill at Creation Only):

– Your job is to summarize the body of an email message (the newest message, if in a thread). Your output should be a briefer version of the message, one paragraph, not bullet points.

– Read the Task Description, which contains an email. Identify the main message body of the email (not including the from, to, and subject lines, nor the signature/sign-off). Exclude any past messages beyond the first, which would begin with subsequent sections of FROM, TO, and Subject.

– Summarize the body message, retaining as much of the original detail as needed to convey the specifics, but make it brief.

– If the task has a comment that is a reply to the original email in the description, review the body of that reply message and include a summary of that as well, separate and identified with a header.

– Use simple wording and language, avoid fluffy language. Don’t reference “the task” just use short-hand to outline key details.

– Always mention users using their names (e.g. John Doe). Never mention them just by their id.

2. AI Field Prompt: Action Items (Custom instructions, Bulleted List, Auto-Updated Only):

– Your job is to review the body of an email message (the newest message, if in a thread) and determine the key action items to take as a result. Your output should be a brief list of bullet points of the action items.

– Read the Task Description and the Email Message custom field. The former contains an email message in full, and the latter contains an extracted body of text comprising the main message.

– With the task description as context, identify any significant action items/next steps to take from the body segment only and list them as individual bullet points.

– If a specific person or user is identified to perform the action item, list their name in parentheses at the end of the bullet. If the action item is clearly for a person on the client side or the client generally, list Client in parentheses instead.

– If the task has a comment that is a reply to the original email in the description, review the body of that reply message and identify any action items obligated by the message. Add bullet points for any action items obligated by that reply.

– Use simple wording and language, avoid fluffy language. If no specific requests or actions to take are prescribed or clearly required, do not try to make one up or suggest broad actions. Only include them if they are specific.

Most of the information we desired we can extract just like this as text in AI fields and display it in columns in the List view. After the email is received in the Client Emails list in ClickUp, these AI fields automatically read the email message and parse out data as described.

AI for action items

 Caption: The Email Manager list in ClickUp, featuring an AI Field for action items. The sidebar shows the prompt setup, guiding the AI to extract and summarize key tasks from each email.

The remaining steps consist of the determining the suggested actions we need to take and who should take them, then creating tasks accordingly. All this requires an AI agent.

Design the AI Agent

This is where our agent take the baton. Here are next steps, all enabled by a ClickUp AI Agent:
  1. Read the auto-generated Action Items field
  2. Create individual subtasks for each action item
  3. Assign each subtask to the appropriate team member in ClickUp, or the Client Team if applicable
  4. Update statuses or closing action items based on comments on the task

 

Creating AI Agents in ClickUp is not much more complicated than creating an AI field. They both utilize natural language prompts and a similar approach. However, agents can contain specific conditions (both logical conditions like regular ClickUp automations, e.g. “When Due Date Arrives”, or natural language conditions like “Only process emails on Monday”) and do much more than fill in a single text field.

 

They can create tasks or documents, change fields on multiple tasks, share or assign tasks with other users, search websites, and analyze some documents/media. By comparison, our goals are light.

 

To start, we went to the AI Agent builder modal and specified our conditions as follows:
  • Only apply to tasks (not subtasks, which could produce an infinite loop)
  • Fire when the task status changes to In Progress
  • Only fire when the task has its Action Items AI field filled

 

This ensures the Agent will only run after the AI fields have already populated, so it won’t try to create tasks without something to reference.

 

We also need to make an intermediary automation to add a tag after the AI Fields are auto-filled. This tag will be used as the actual trigger point to fire the agent here. You could imagine many other triggers, such as every morning at 9am, feel free to adjust. Regardless, this step is very simple as every AI field creates its own automation. We just need to add one action as below:

set custom field, screenshot graphic

Next up, writing the prompt language itself. As we have outlined the actions we need the agent to take, the prompt language flows linearly from the list above. AI experts the world over love to begin by describing what “job” or “role” the agent is to perform, to contextualize the work. Describe that role briefly, its goals and any conditions, at the top. Then indicate in a clear, linear order, the steps to proceed. For each step, indicate if there are any prerequisites or fields to reference — and be sure to insert direct links to any tasks, docs, people, etc. to be as explicit as possible.

 

Formatting the entire prompt as a list is popular and makes revision much simpler. (Although the final two are specific to the ChatGPT API, OpenAI’s best practices for prompt writing are highly useful:

AI agent

After a few tests and edits, we landed on the prompt language below. The initial listed text contains the main next steps prescribed above, while the text after regards a more advanced automation discussed a bit later.

AI Agent Prompt: Auto-Create Subtasks for Action Items

– Your job is to read a text field of suggested action items and assignments, then create each action item as a subtask and assign it to the right team member.

  1. Read the action items listed in the Action Items AI field of the task. Each action item is a single bullet and contains a suggested assignee at the end, named in parentheses.
  2. For any items where the suggested assignee’s name matches a ClickUp user, create a subtask (if there isn’t one matching that action item already). Set the subtask name to match the text of the action item, except the name in parentheses.
  3. Assign the subtask to the user matching the name in parentheses in the action item. If the name says Client, assign it to the team Client Team.

-Subtasks should remain open if it’s task name matches an action item listed in the Action Items field, and should be closed if it does not.

-If there are other subtasks open under this task, check if the last comment mentions they were completed or done, and if so, mark that subtask as done.

Ai agent extraction screenshot graphic

 Caption: The AI Agent extracts action items from the email, populates them in a custom field, and automatically creates and assigns subtasks based on those actions—streamlining follow-up and accountability.

Now, we have a functioning AI Agent that follows all of the aforementioned steps to turn our suggested actions into assigned subtasks. When the Action Items field is filled, the task is marked in Progress, and the Agent goes to work. It may take a minute to populate everything as expected, depending on the complexity of the email.

 

And important caveat: these subtasks cannot be assigned to users who do not already have access to the task or list. If the list is not shared with everyone relevant, you may consider invoking steps in the AI Agent prompt (or even creating another agent) to share tasks to the suggested assignee first, then assign it.

 

 

Now Iterate and Refine the System

After reviewing the formatting of the outputs, we realized some minor adjustments were necessary. Referencing a ClickUp Doc or a formatted paragraph as an example in the prompt ensures the AI Agent will closely match your desired output. I adjusted some destination fields from a default “summary” type bulleted list to a more natural language paragraph style.

 

Extracting action items, identifying assignees, and assigning subtasks required significant testing and tweaking. Mainly though, we just wanted to make it do more. As we intimated early, a more advanced step you can include is to prompt the agent to monitor comments for progress and update subtasks accordingly.

 

Using another automation, we can trigger the Action Items AI Field to auto-regenerate when a comment is added. The field will then reference the original message and comment activity and add or remove anything as needed.

reiterate system, screenshot graphic

Then, when a user comments that a new action item is required, or a prior item is completed, the AI Field will update. The Agent will then run again to create or close subtasks automatically! This keeps your task progress tracked with a simple comment-based update. 

 

This layered, multi-step approach means you’re not just parsing data and checking boxes, you have a trackable workflow you can watch as work progresses.

 

 

Finally, Monitor and Maintain

In this relatively simple case, we’re producing text summaries and suggested actions, then assigned tasks. There’s a possibility some unfinished work might be closed prematurely with the advanced automation step above, or potentially assigned to the wrong person, or not assigned. All of these potential gaps require testing and validation before deploying this solution at scale. For a small team, you can typically troubleshoot as you use the agent and refine it incrementally.

 

In this case here, you can occasionally compare the email summaries to the original message in the task description, and the action items field to the corresponding tasks. If you find a summary or set of action items is inaccurate, or the estimated time to perform them is incorrect, continue refining the prompt.

 

How to Maintain Your ClickUp AI Agent

  • Ensure emails are properly forwarded: check your inbox and outbox where the rule has been set up regularly for any misses.
  • Spot-check email bodies and agent-generated summaries to confirm no key details are missed.
  • Confirm subtask assignments look appropriate.
  • Optionally, you can create another agent to run a side-by-side comparison and suggest tweaks to the prompt (meta-metaprompting?), but always supplement this with human eyes.

 

For more sensitive cases, for instance where the Agent is either taking multiple/cascading actions or producing some more critical data point, it’s critical to review the Agent activity and audit trail on a regular basis.

 

This is accessible via the Automation/Agent manager window under the Activity tab. Agents that misfire, skip tasks inappropriately, or produce unreliable results can cause havoc. For this reason, we always validate our configurations with rigorous testing and a written audit process.

 

Knowing the expected outcomes of each task, step by step, and comparing them to test tasks periodically, will help mitigate issues early.

 

In the End

Automating email management in ClickUp with AI Agents and AI Fields is one big piece of the puzzle to creating a single source of truth for your operations. You can use this approach to create multiple agents and transform your team’s scattered requests into a truly manageable operations hub.

 

Supercharge ClickUp’s native features with robust AI-driven analysis, task management, and decision-making, ensure no request slips through the cracks. If you want to save time and keep your team focused on what matters most with ClickUp AI, book a meeting with the ClickUp AI experts at Tuck.

Ruby Partners

As official Ruby partners for ClickUp, we are licensed to provide you with ClickUp branded professional services. This also makes us, and our consultants, ClickUp experts.

Alex Morgan

Alex Morgan

Professional Services Practice and Project Management Professional

Alex Morgan (he/him/his) is the IT Professional Services Practice Lead at Tuck Consulting Group. In 2021, Alex began with Tuck as a pro bono consultant for local nonprofits and instantly clicked with the company culture, mission, and clients. Working with diverse clients in tech and adjacent sectors has afforded a breadth of relationships and invaluable strategic insight.

Alex works remotely from Burlington, Vermont. He is a certified ClickUp Expert and is certified with the HubSpot Content Management System and HubSpot Marketing Hub.

When work is over for the day, Alex enjoys live music, NBA basketball, and critiquing TV shows with his daughter (we are equally intrigued to hear more about this last one).

Keep Reading

Related Articles

Share This