Your Stuff, Across The Platform
Your bookmarks, drafts, open tasks, and inbox — found and triaged from chat, hard-scoped to you. It can bookmark, edit a note, save a private note, or clear notifications once an admin allows it.
Why Your Own Stuff Is The Hardest Thing To Find In A Big Platform
The platform has tens of thousands of records. Yours are a handful — but they're scattered across workspaces, wikis, tasks, the inbox, and a dozen apps. Personal Agent answers "where's my stuff?" without making you click through every surface.
Your Bookmarks Live In Ten Different Apps
You bookmarked a wiki, a workspace message, a recognition post, a recruiting job, and an SOP — and you remember none of them. The platform has the data; finding it means visiting each app's bookmark list one by one. Most people just re-find the thing from scratch.
Half-Written Drafts Disappear Forever
You started a workspace message yesterday, got pulled into a call, and forgot it. Today the thread moved on without you. Drafts exist — the platform autosaved them — but there's no single "show me everything I haven't finished" view.
"My Tasks" Means Tasks From Five Different Apps
A task from a workspace, a task from onboarding, a task from a service-desk ticket, a task from a workflow — they all land in different inboxes. "What's overdue?" is a five-page click tour.
The Notification Inbox Becomes Wallpaper
230 unread. Most don't matter. Three of them are the @-mentions you actually needed to see before 9am. The inbox is so noisy that triaging it manually is its own job.
"What Was I Working On?" After A Two-Week Vacation Is A Half-Day Of Spelunking
Back from leave, you open the laptop and have no idea where you left off. The last doc you edited, the workspace you were active in, the threads waiting on your reply — they're all there, scattered across six apps. A "what's mine and recent?" rollup would save the morning; instead you reconstruct the picture tab by tab.
Your Own Profile Data Is Hard To Find In Your Own Tool
"What's my employee ID?" "Where's my offer letter filed?" "Who's my listed emergency contact?" Each lives in a different app's profile surface, none of them indexed together. A scoped chat answer would take three seconds; the actual lookup takes three minutes of clicking through menu items you only use once a year.
Personal Agent At A Glance
Personal AI
Searches your own bookmarks, notes, drafts, tasks, inbox, shifts, leave, and pay. Hard-scoped to your account.
Inside Personal Agent — The Actual Capabilities
Every block below maps to a real tool the agent uses against your own data. Reads run against your account only; the write actions stay off until an admin enables them, and the writes_enabled toggle turns them on or off per tenant.
Find Anything You've Bookmarked, Across Every App
Your bookmarks span eleven different record types — workspace messages, comments, tasks, wikis, HR policies, SOPs, library items, site pages, recognition posts, recruiting jobs, app builder records. The agent searches all of them by title and by note text, so you can find the thing you saved even when you've forgotten what app it lived in.
- Search across every bookmarkable type — search_my_bookmarks covers WorkspaceMessage, WorkspaceMessageComment, Task, Wiki, HrPolicy, Sop, LibraryItem, SitePage, RecognitionPost, RecruitingJob, PluginRecord.
- Substring match against title AND note — find a bookmark by the reason you saved it, not just the title of the record.
- Filter by type — "show me my bookmarked wikis" / "my bookmarked SOPs" without scrolling through everything.
- Limit defaults to 25 — the agent returns a tight ranked list, not a wall of every bookmark you've ever made.
Pick Up The Drafts You Left Half-Written
Messages, comments, broadcasts, communications and tasks you started in the platform's composers are autosaved as drafts. The agent surfaces them in one list, linked back to the composer where it can reopen them, so the half-written reply from yesterday doesn't get buried.
- {"List every unfinished draft — list_my_drafts covers drafts saved through the platform composers" => "workspace messages and comments, to-dos, direct messages, broadcasts, communications, tasks, notes and app drafts. Apps that keep their own drafts, such as Surveys and Forms, are not included."}
- Filter by kind — ask for just your broadcast, communication or workspace-message drafts when you only care about one.
- Last-edited timestamp on every row — easy to tell the draft from this morning apart from the one from three weeks ago.
- Personal-only by construction — you only ever see your own drafts; never another user's.
Triage Your Tasks Without Touching Five Apps
Tasks land from a dozen sources — workspace assignments, onboarding flows, service desk tickets, performance reviews, surveys. The agent filters them down to open, overdue, or "due in the next N days" so you can plan the day from one prompt.
- List tasks assigned to you — list_my_tasks filterable by status (open / completed / all / overdue).
- {"Due window filter — due_within_days" => "7 for \"what's on my plate this week\"."}
- Overdue surfaced explicitly — "show me what's overdue" returns a ranked list, not a search you have to interpret.
- Scoped to your account — never returns tasks assigned to anyone else, even if you'd nominally have access to view them.
Summarize Your Inbox Without Reading 200 Notifications
The notification inbox builds up fast — @-mentions, task assignments, ack reminders, workspace updates. The agent summarizes the unread set with counts plus the most recent handful, so you can see what matters without scrolling.
- Summarize your inbox — summarize_my_inbox returns unread counts plus the most recent N items (default 10), or every open request up to 50 when you ask what needs your action.
- Unread-only by default — the noise from already-read items is filtered before the summary runs.
- Action-required count — the summary counts the open requests still waiting on you (read or unread) and can list just those, so they don't get lost among the newest items.
- Personal scope — the agent never reads anyone else's notification inbox.
Write Actions, Off Until You Turn Them On
The agent can save a bookmark, remove one, update a bookmark's personal note, save a private note, mark notifications read, or acknowledge a must-read post or broadcast. Every one of those touches only the asking user's own data, so all but acknowledging run without a per-action confirmation prompt (acknowledging shows a card naming the item first) — the gates are structural instead: the tenant-level writes_enabled toggle ships OFF, and a per-user daily cap bounds them once it is on. With writes off, they are removed from the model's tool list and refused server-side.
- Writes — bookmark_record, remove_bookmark, update_bookmark_note, mark_inbox_read, create_my_note, confirm_item (asks first), decide_waiting_item (asks first) — all off by default, all bounded by a per-user daily cap (50/day by default, configurable 0-1000, where 0 means unlimited).
- Bookmark any of 11 record types with an optional 500-char personal note.
- {"Mark inbox read selectively or in bulk — pass notification_ids for surgical clears or all" => "true to clear everything visible in the inbox (snoozed items stay unread), both bounded by the same daily cap."}
- writes_enabled tenant toggle — when off, the write tools are removed from the model's tool list and refused server-side.
Outcomes Teams Can Measure
Personal Agent isn't a customer-facing metric mover — it's a productivity backstop for every user. The right measurements are about adoption and the day-to-day fragmentation it absorbs, not about deflection or cycle time.
- Daily active users hitting Personal Agent at least once for bookmarks / tasks / drafts / inbox.
- Time-to-first-action on overdue tasks — median minutes from "what's overdue?" to a status change.
- Inbox triage rate — share of unread notifications cleared via the agent vs the standalone inbox UI.
- Bookmark recall — share of bookmarks retrieved via search vs the standalone bookmarks page (signals whether users can actually find the things they saved).
- Write-toggle adoption — share of tenants that turn writes_enabled on after starting with writes off.
Hard-Scoped To The Asking User · No Cross-Account Reads, Ever
Personal Agent's tools look up only the asking user's own records — there is no parameter for another user's bookmarks, notes, drafts, tasks, inbox, shifts, leave, pay, or certifications, not even with admin elevation. Beyond that it searches only company knowledge the user can already open. The writes touch only that user's own data, ship disabled, and are bounded by a per-user daily cap.
- No parameter for another user — every personal lookup filters to the asking user.
- 7 writes, all off by default — bookmark_record, remove_bookmark, update_bookmark_note, mark_inbox_read, create_my_note, confirm_item, decide_waiting_item.
- writes_enabled tenant toggle — when off, the write tools are removed from the model's tool list and refused server-side.
- Audit trail on every tool that runs — each call to the agent's own tools that executes records the requesting user, the tool, and its argument names (never values); each write that changes something is also logged in AI Operations with the ids it was given (never note text).
WHAT TEAMS TRY INSTEAD
The four alternatives — and why none of them can see your bookmarks, notes, drafts, tasks, inbox, and shifts at the same time
Most employees default to one of these before they discover Personal Agent. None of them combine cross-app reach with a per-user identity scope that the platform enforces.
Pasting links into ChatGPT, Claude, or Copilot
General-purpose AI doing one-shot tasks on copied text
- Personal Agent already knows your bookmarks, notes, drafts, tasks, inbox, shifts, leave, and pay — nothing to paste
- Acts on records (mark inbox read, edit a bookmark note) instead of producing prose about them
- Stays inside the platform's permission model so the agent literally cannot list another user's data
Microsoft 365 Copilot / Google Workspace Duet AI
Vendor-trapped personal AI that ends at the suite boundary
- Reads your own shifts, leave balances, pay periods and expiring certifications live from the apps that own them — not just mail and docs
- Works for the deskless and frontline users who never had M365 or Workspace seats in the first place
- user_id = current_user.id is hard-scoped at the tool layer, not an opt-in security mode
Custom personal assistant build on the intranet
An engineering team's roadmap item that never quite ships
- Shipped already with bookmarks, notes, drafts, tasks, inbox, shift, leave, pay, and certification reads on day one — no integration sprint
- 7 admin-gated writes covered (bookmark_record, remove_bookmark, update_bookmark_note, mark_inbox_read, create_my_note, confirm_item, decide_waiting_item) with audit logging
- Cross-app reads are a fixed set — shifts, leave, pay and certifications — each answered by the app that owns it
The manual fallback — find it yourself later
Browser tabs, bookmark folders, and a personal triage routine that loses to the inbox
- Where is that bookmark I saved last week — answered in one prompt instead of scrolling
- Drafts you forgot you started come back into view in one list, linked to the composer that can reopen them
- Inbox triage moves from scroll-and-pray to show-me-what-is-actionable
PLATFORM ADVANTAGE
Personal Agent inherits everything the platform already enforces
A standalone personal assistant has to plumb identity, permissions, audit, and mobile delivery itself. Personal Agent gets all of it for free.
Hard-scoped identity
Every personal lookup is filtered to the asking user inside the tool itself, and no tool accepts another user — by design.
Cross-app data plane
Reads your bookmarks, notes, drafts, tasks, and notifications, plus your own shifts, leave balance, paychecks, and expiring certifications — each answered live by the app that owns it, when that app is enabled.
Admin-gated, daily-capped writes
7 writes (bookmark, unbookmark, edit note, save a note, mark read, acknowledge a must-read, decide a request waiting on you) ship OFF. Once an admin enables them they run immediately — the first five touch only the asking user's own data; acknowledging and deciding show a card naming the item first — bounded by a per-user daily cap and logged to the audit trail.
writes_enabled tenant toggle
Admins can deploy with the write actions off. Until the tenant opts in, those tools are removed from the model's tool list and refused server-side.
Audit trail on every tool that runs
The platform's AI tool log records the requesting user, the tool, and its argument names (never values) for each call to the agent's own tools that runs, and AI Operations lists each write that changed something with the ids it was given.
Built for the frontline
Lives inside Ask AI in the same mobile app frontline employees already opened for shifts and pay — no separate sign-in, no separate app to install.
INDUSTRY FIT
Where a per-user agent moves the needle
Personal Agent helps everywhere, but the lift is biggest in industries where employees touch many apps a day and have no desk to organize them from.
Retail
Store associates triage shift, training, and recognition notifications between transactions instead of after-shift on a personal device.
Healthcare
Clinicians find the policy bookmark from last month's safety in-service in one prompt instead of scrolling through a charge nurse's folder.
Manufacturing
Plant-floor crews on shared kiosks see only their own tasks, drafts, and inbox — never the previous user's — because identity is hard-scoped at the tool layer.
Hospitality
Property staff pick up unfinished maintenance drafts and overdue tasks at the start of every shift without bouncing across four apps.
Financial Services
Bankers and advisors mark compliance notifications read, and each change is logged in AI Operations with who asked, the tool, and — when they named specific notifications — those ids.
Public Sector
Each employee's questions reach only their own records, write actions can stay switched off, and AI requests are processed by the platform's configured AI model provider, like every MangoApps AI feature.
WHY MANGOAPPS WINS
An embedded personal agent beats a horizontal AI, a suite-trapped Copilot, or a custom build on every axis
The argument employees, IT, security, and finance all share — and the one ChatGPT or Microsoft Copilot structurally cannot answer.
Cheaper than the alternatives
No per-seat ChatGPT Plus license, no Microsoft 365 Copilot add-on at $30/user/month, no engineering team building yet another internal assistant.
More secure
Every personal lookup filters to the asking user, write actions are admin-gated and daily-capped, and each write that changes something is logged in AI Operations. AI requests go to the platform's configured model providers, like every MangoApps agent.
Easier to deploy
Nothing to build or integrate — an admin enables the licensed app from the Apps Marketplace and it answers inside Ask AI. Toggle writes_enabled when you want write tools live. No identity provider rewire, no separate vendor onboarding.
Easier to use
Lives inside the Ask AI sidebar across every page in the platform. Frontline employees use it on the same mobile app they already opened for the day.
Easier to manage
One toggle to enable writes, one configuration screen, one audit log. Per-business preferences sit beside every other app's settings.
Easier to extend
Shares the agentic-tool framework with every other MangoApps agent. New personal capabilities ship as tools, not as a rebuild.
AI is actually better
Copilot can summarize your inbox. Personal Agent also lists your unfinished workspace drafts, edits a bookmark note from last quarter, and surfaces your overdue tasks across every app — all from one prompt, all hard-scoped to you.
At our head office within HR, marketing, and finance, we use ChatGPT quite a bit, and MangoApps AI Helper is a useful tool for us. As I go around to the stores, I’ve been showing them how the AI Intranet Helper can find things and answer plain language questions about our resources, and they’re quite excited about that.
Alison van Zijl
Head of Human Resources
Full House
Customer Success
MangoApps Customers
Organizations that trust MangoApps to run every workflow across their workforce.
Customer Case Studies
Customer Case Studies
Customer Case Studies
Video Case Studies
Customer Case Studies
Customer Case Studies
Frequently Asked Questions About Personal Agent
16 tools — 9 read and 7 write. Five reads (search_my_bookmarks, search_my_notes, list_my_drafts, list_my_tasks, summarize_my_inbox) surface your saved items, private notes, half-written drafts, open/overdue tasks, and unread notifications. Four more (my_upcoming_shifts, my_leave_balance, my_recent_paychecks, my_expiring_credentials) are answered live and read-only by the apps that own them, and disappear when those apps aren't enabled. Writes (bookmark_record, remove_bookmark, update_bookmark_note, mark_inbox_read, create_my_note, confirm_item, decide_waiting_item) are gated by a tenant-level writes_enabled toggle and a per-user daily cap.
No. Personal Agent's tools do not accept another user — every lookup is filtered to the person asking, and admin elevation doesn't change that.
Eleven — WorkspaceMessage, WorkspaceMessageComment, Task, Wiki, HrPolicy, Sop, LibraryItem, SitePage, RecognitionPost, RecruitingJob, PluginRecord (App Builder records). Each bookmark can carry an optional 500-character personal note that the agent can search and edit.
By two structural gates rather than a per-action prompt. Because every Personal Agent write changes only the asking user's own data — their bookmark, their note, their read-state — the tools run as soon as the user asks, with no "confirm?" step. The one exception is acknowledging a must-read post or broadcast: an acknowledgement is compliance evidence that cannot be taken back, so a card naming the item asks first. What bounds them is the tenant writes_enabled toggle, which ships OFF and is the kill switch (when off the write tools are removed from the model's tool list and refused server-side), and a per-user daily cap of 50 actions, configurable from 0 to 1000, where 0 means unlimited. Every successful write that changes something is logged with the requesting user, the tool and the ids it was given — a bookmark, a note, or the notifications named (a mark-all or mark-by-topic records only how many were marked; never note text) — and admins see cap usage and adoption on the app's Analytics tab. One nuance worth knowing: removing a bookmark that carries a note sends it to Trash for 30-day recovery, while one without a note is deleted outright.
Daily active users hitting Personal Agent, time-to-first-action on overdue tasks, inbox triage rate via the agent vs the inbox UI, bookmark recall (search vs standalone page), and the share of tenants that flip writes_enabled on after starting with writes off.
Let's Talk
Every team. Every employee. Every workflow. One AI platform — built for the frontline. Since 2008, trusted by 2 million+ users with an NPS of 78.
Why Choose Us?
- Frontline AI: Governed AI for every employee and workflow.
- Top Security: HITRUST, ISO & SOC 2 certified.
- Exceptional UX: Delightful on mobile and desktop.
- Proven Results: 98% customer retention rate.
Rolled out to every employee at AutoZone (125,000), PetSmart (50,000+), A.S. Watson and Raley's (20,000) — and at larger retailers we are not permitted to name.
Prefer to explore first? Ask AI about Personal AI Agent →