Think about the last deadline you missed. Was it a task sitting in Jira that you ignored? Probably not.
More likely, it was something buried in a Teams message: "Hey, can you send me those figures before the client call?" Or an email you replied to with "Sure, I'll have that to you by Friday" - then promptly forgot.
These aren't tasks in your system. They're invisible commitments scattered across your communication tools. And they're the real reason things fall through the cracks.
The Invisible Task Problem
Here's a stat that should worry you: Research suggests that 60% of workplace commitments are made in communication tools, not project management systems.
Think about where work actually gets assigned:
- A manager asks for something in a Teams channel
- A client requests updates via email
- A colleague @mentions you in a group chat
- Someone replies to a thread with "Can you handle this?"
None of these automatically become tasks. They exist only as messages - ephemeral, unsearchable, and easy to lose in the scroll.
The @Mention Trap
Teams and Slack @mentions are particularly dangerous. They feel urgent when they arrive. You see the notification, you read the message, you think "I'll do that later."
But "later" never comes with a reminder. The mention scrolls away. By tomorrow, it's buried under 50 new messages. By next week, it might as well not exist.
And yet, the person who sent it absolutely expects you to follow through.
Email Commitments Are Even Worse
At least with chat, messages are somewhat visible. Email commitments are truly invisible.
You write: "I'll review this and get back to you by Wednesday."
Where does that commitment live? In your Sent folder. A place you never check. There's no task, no reminder, no system tracking your promise.
Wednesday arrives. The person follows up: "Did you get a chance to review this?"
You didn't. Because reviewing it was never a task - it was just something you said.
AI-Powered Task Extraction
What if your communication tools could automatically identify when you've committed to something?
This is exactly what NextUp does. Using AI, it scans your Teams messages and emails to detect:
- Direct requests: "Can you send me the report?"
- Commitments you've made: "I'll have that ready by EOD"
- Follow-up triggers: "Let me know when you've reviewed this"
- Deadline mentions: "This needs to be done before the board meeting"
- @Mentions requiring action: Not just FYI mentions, but ones expecting a response or deliverable
The AI filters out noise - newsletters, automated notifications, casual conversation - and surfaces only messages that represent actual work.
How It Actually Works
Here's a realistic example.
You receive this Teams message from your manager:
"@YourName - the Henderson proposal needs final numbers before Thursday's pitch. Can you pull the Q4 actuals and send them to Sarah? She's building the deck."
NextUp's AI extracts:
- Task: Pull Q4 actuals and send to Sarah
- Context: Henderson proposal
- Deadline: Before Thursday
- Related person: Sarah (building the deck)
This becomes a tracked task in your unified view - not buried in Teams, but sitting alongside your Jira tickets, Asana tasks, and calendar commitments.
Thursday morning, if you haven't marked it complete, you'll see it flagged as urgent.
Email Extraction Works Similarly
You reply to a client email:
"Thanks for sending this over. I'll review the contract and send my comments by Monday."
NextUp scans your sent emails and identifies:
- Commitment: Review contract and send comments
- Deadline: Monday
- Context: Client name, email thread
Now your promise is tracked. Monday morning, you'll see "Review contract for [Client]" in your task list - even though you never manually created a task.
What Gets Captured (and What Doesn't)
The AI isn't magic - it's trained to identify specific patterns that indicate actionable work.
Captured as Tasks:
- Requests with clear deliverables ("Can you prepare...", "Please send...")
- Your commitments with deadlines ("I'll have it done by...")
- Approval requests ("This needs your sign-off before...")
- Action items from meeting summaries ("Tom to follow up on...")
- Explicit @mentions asking for something
Filtered Out:
- Newsletters and marketing emails
- Automated notifications (build alerts, calendar invites)
- FYI messages with no action required
- Casual conversation ("How was your weekend?")
- Acknowledgments ("Thanks!", "Got it")
You can also configure VIP senders whose messages always create tasks, and blocked senders whose messages never do.
The Difference This Makes
Consider your typical Monday morning.
Without AI extraction:
You open Teams. 73 unread messages across various channels. You scroll through, trying to remember if anyone asked you for something. You check a few threads. You probably miss something. You open email. 47 new messages. You skim subject lines. A few look important, most don't. You open your task manager. It shows the tasks you manually created - none of the commitments hiding in your inbox.
With AI extraction:
You open NextUp. Your unified task list shows:
- 4 AI-extracted tasks from Teams (including one from Friday you'd forgotten)
- 2 commitments you made in email last week
- Your existing Jira tickets and Asana tasks
- Today's calendar with available time calculated
Everything that needs your attention, in one place. Nothing hidden in message threads.
Common Objections
"Won't this create too many tasks?"
The AI is conservative by design. It's looking for clear action items, not every message that mentions your name. In testing, users report 3-5 extracted tasks per day from email/Teams combined - not overwhelming, but enough to catch the commitments that would otherwise be lost.
"What about privacy?"
NextUp uses OAuth for email and Teams access - it never stores your passwords. Message content isn't saved; only the extracted task details are retained. When you revoke access, all associated data is deleted.
"I'd rather just be more disciplined about creating tasks manually"
You could. But will you? Every time someone asks you for something in chat, will you stop, open your task manager, and create a task? Every time you make a commitment in email, will you add a reminder?
Most people don't. Not because they're lazy, but because the friction is too high. AI extraction removes that friction - commitments become tasks automatically.
Setting Up AI Extraction in NextUp
Getting started takes about 5 minutes:
1. Connect Your Accounts
Authorize NextUp to access your Microsoft 365 (for Outlook and Teams) or Google Workspace (for Gmail). This uses standard OAuth - your credentials stay with Microsoft/Google.
2. Configure Extraction Settings
Choose which accounts to scan, set a lookback window (typically 7-14 days), and optionally add VIP/blocked senders.
3. Review Extracted Tasks
Initially, review what the AI captures. Mark false positives as "not a task" to improve accuracy. After a week, the system learns your patterns.
4. Integrate with Your Workflow
Extracted tasks appear alongside your other work. Triage them like any other task - complete, snooze, or dismiss.
The Bigger Picture
AI task extraction isn't about replacing human judgment - it's about ensuring nothing slips through the cracks.
You'll still prioritize your own work. You'll still decide what matters. But you'll do it with full visibility into everything that's been asked of you, not just the subset that made it into your formal task systems.
Because the most dangerous tasks aren't the ones you're avoiding. They're the ones you've already forgotten.
Stop losing tasks in your inbox. NextUp uses AI to automatically capture action items from Teams, email, and chat - so commitments become tracked tasks, not forgotten messages.
Try NextUp Free | No credit card required
Tom Foster is the founder of Avoidable Apps, building tools that eliminate the busy work fragmenting modern knowledge workers' attention.