Personal Agent FAQ
Answers to common setup and operating questions about Personal Agent.
For what the app is and how to set it up from scratch, see the
Personal Agent Overview.
Setup
Do I need to do anything before users can talk to the Personal Agent?
Yes — two things. First, the app must be licensed and enabled for your tenant in the Apps Marketplace — the AI Productivity suite and All Employee Suites include it. Second, the Enable Personal Agent toggle under Apps → Personal Agent → Settings must be on. Both must be true before the agent responds to personal-data questions through Ask AI.
What’s the difference between enabling the app and enabling the agent toggle?
Enabling the app in the Apps Marketplace turns the app on, and the agent with it — Enable Personal Agent starts switched on. The toggle is not a separate go-live step: turning it off stops the agent and also turns the app off for your whole organization, so unless Personal Agent comes with your plan, employees lose the Overview and My Activity pages too. Enabling the app again from the Apps Marketplace brings the pages back and turns the agent back on with them.
How do I test the agent before rolling it out?
On the Settings page, use the Generate sample data button. It creates a small set of drafts, notifications, and bookmarks scoped to your own account, tagged so they can be removed cleanly. Then open Ask AI and try a prompt like “What’s in my inbox?” to confirm the agent responds.
Permissions and access
Who can change Personal Agent settings?
Any user with an admin or above role, or a user designated as a Personal Agent app admin. Two things stay with business admins: only a business admin can switch Enable Personal Agent on or off, and only a business admin’s Export CSV includes user IDs — a Personal Agent app admin’s export replaces them with anonymous keys. The Overview and My Activity tabs are visible to all users; Settings and Analytics are restricted to those roles.
Can the agent see other users’ data?
No. Every query is filtered to the asking user’s ID. No tool accepts a target user parameter. If someone asks for another user’s bookmarks, the agent explains it can only see the asking user’s own data.
Day-to-day
What does “Allow write actions” control?
When off (the default), the agent cannot bookmark or un-bookmark records, edit bookmark notes, write private notes, or mark notifications read — it can still search and list them. When on, the agent gains five write tools: bookmark a record, remove a bookmark, update a bookmark’s note, write a private note, and mark notifications as read. Each write runs immediately without a confirmation prompt because every action changes only the user’s own data; the admin toggle and the Per-user daily write cap are the safety limits. Not every write can be undone: removing a bookmark that carries a note sends it to Trash for 30-day recovery, one without a note is deleted outright, and notifications the agent marks read can be marked unread again from the inbox. This switch does not cover features Ask AI offers in every conversation — tracking a Journey, forgetting a remembered fact, or (for admins) starting an App Builder draft — which have their own settings.
What types of records can users bookmark through the agent?
The agent supports 11 record types: workspace messages, workspace comments, tasks, wiki pages, HR policies, SOPs, library items, site pages, recognition posts, job listings, and app builder records. These are the same types available through the bookmark icon in the UI. Before creating a bookmark, the agent checks that the user can actually see the target record.
How does the per-user daily write cap work?
Every write that actually changes something counts against the user’s daily limit — re-bookmarking a record that is already saved, or marking already-read notifications read, does not. The default is 50 writes per user per day, configurable from 0 to 1,000 in Settings. Setting it to 0 means unlimited. The cap resets at midnight in each user’s own time zone (their location’s, otherwise their profile’s), not on a rolling 24 hours. When a user hits the limit, the agent tells them and stops accepting writes until the next day. Reads are never capped.
What happens when someone asks the agent to mark their inbox as read?
The agent marks all visible unread notifications as read. Snoozed items are deliberately excluded — they stay unread so they resurface at their scheduled time. Users can also ask the agent to mark specific notifications, or every visible notification whose title mentions a topic (“mark my inspection notifications as read”), rather than clearing everything.
How do users check their own agent usage?
Any user can go to Apps → Personal Agent → My Activity to see their tool-call history: total calls, success rate, which tools they have used (labelled read or write), and their 10 most recent calls. This view is also available on mobile. Admins see cross-user metrics under the Analytics tab.
Cross-app lookups
What personal data from other apps can the agent access?
When the source apps are enabled for the tenant, the agent can pull the user’s own data from four apps: Shifts & Scheduling (upcoming shifts), Leave Management (leave balances), Payroll Prep (recent pay periods), and Skills & Credentials (expiring certifications). These lookups are read-only and follow the same privacy scope — only the asking user’s records.
A user asked about their shifts, leave, pay or certifications, but the agent said it can’t check — why?
The agent says which of these applies:
- The source app isn’t available to that user — Shifts & Scheduling, Leave Management, Payroll Prep or Skills & Credentials is not enabled for your tenant, or that user isn’t allowed to use it. An admin can enable the app in the Apps Marketplace.
- The app’s own AI agent is switched off in that app’s settings (AI Agent Enabled, or Enable Payroll Agent in Payroll Prep). The app itself is on; turn its agent back on there.
- That lookup is turned off inside the app — Team Calendar in Shifts & Scheduling (upcoming shifts), Balance Tracking in Leave Management (leave balances), or Certification Management in Skills & Credentials (expiring certifications).
- Upcoming shifts only: Enable Schedule Queries is turned off in Ask AI’s settings.
- The app couldn’t be reached just now — the agent says so and asks the user to try again in a moment; nothing needs to be switched on.
When something looks wrong
“I searched my bookmarks but some results are missing”
Bookmark text search scans the most recent 2,000 bookmarks. If the user has more than 2,000 and the missing item is older, the agent adds a note suggesting a record-type filter to narrow the search. Filtering by type searches within that type’s 2,000 window, which usually surfaces the result.
Bookmarks whose record has since been deleted, or that the user can no longer open (for example after leaving a private workspace), are also left out of the agent’s answers. The Bookmarks page still lists them — marked “This item was deleted or is in the trash” or “no longer available to you” — with a Remove from bookmarks link.
“The agent said it couldn’t find a record I asked it to bookmark”
The agent checks that the user can see a record before bookmarking it. If the user does not have access — because of permissions, audience scoping, or the record being in a private workspace — the agent responds with “not found” without revealing whether the record exists.
“Write actions are on but the agent says they’re disabled”
Both the Enable Personal Agent toggle and the Allow write actions toggle must be on, and the user must not have exceeded their daily write cap. If the agent toggle is off, writes are blocked even if the write-actions toggle is on. Check all three conditions in Settings.
“My employee says Personal Agent doesn’t appear in Ask AI”
Confirm the app is licensed and enabled for your tenant in the Apps Marketplace (the AI Productivity suite and All Employee Suites include it), that it is published to that employee, that they have Ask AI, and that Enable Personal Agent is on in Settings.
Employee-reported questions
“How do I stop the AI from seeing my personal data?”
There is no per-person opt-out today. Personal Agent reads only the asking person’s own bookmarks, notes, drafts, tasks and inbox — never anyone else’s — and only when that person asks it something, in Ask AI or an AI assistant they connected. An administrator decides whether Personal Agent is available for the organization and who it is published to.
“I snoozed some notifications and then asked the agent to mark everything as read — did it clear them?”
No. When the agent marks all notifications as read, it only touches items currently visible in the inbox. Snoozed items stay unread so they come back when their snooze period ends.
Licensing and limits
Are there limits I should know about?
| Limit | Value |
|---|---|
| Bookmark note length | 500 characters |
| Private note title length | 200 characters |
| Bookmark text search window | 2,000 most recent bookmarks |
| Per-user daily write cap | Default 50 (configurable 0–1,000) |
| Analytics time ranges | 7, 30, or 90 days |
| Analytics CSV export | 50,000 rows maximum; user IDs for business admins, anonymous keys for Personal Agent app admins |
| Supported bookmarkable record types | 11 |
| Included in | AI Productivity, All Employee Suites |
More help
- Personal Agent Overview
- Ask AI — the assistant answers Personal Agent questions from these articles.