Ask AI App Overview

Ask AI

1. What it is

Ask AI is the platform’s conversational assistant. Employees type or speak a question, and Ask AI routes it to whichever specialist agent owns the answer — scheduling for shift questions, leave management for PTO balances, service desk for ticket status, and so on. Each agent works with that app’s live data, so answers reflect the current state of a schedule or queue, not just what a policy document says.

• Enablement: Ask AI requires a licence and the Platform Standard edition or above. An admin enables it from Admin → Apps → All Apps → Ask AI, and can restrict access to specific groups or roles.
• What it is not: Ask AI is not the place to pause or resume the AI across the organisation — those kill switches live in Admin → AI Operations.

2. Standing it up

1. Go to Admin → Apps → All Apps, find Ask AI, and enable it. Confirm the organisation is on the Platform Standard edition or above — Ask AI will not load on Platform Lite.
2. Open Apps → Ask AI → Admin → Settings and review the defaults. The key decisions at this stage:
   • Enable Action Execution (on by default) — whether the assistant can carry out actions such as filing tickets or creating shifts. Changes like these pause for the employee’s confirmation first; only small changes to their own things, such as a bookmark or a page watch, run straight away.
   • Enable Voice Mode (on by default) — whether employees can speak to Ask AI. The default daily limit is 30 minutes per user.
   • Conversation retention (default 24 months) — how long transcripts are kept before the nightly sweep permanently deletes them.
3. Open Apps → Ask AI → Admin → Agent Guidelines → AI Features and confirm the toggles suit your rollout. Key defaults:
   • Multi-agent answers is on — a question that spans several areas is sent to each relevant agent, and their answers are combined into one reply.
   • Shared knowledge search is on — agents also search the Company knowledge sources ticked beneath it.
   • Review that source list. Library items may be off if your organisation already had libraries when this source was added. Tick it to let answers use your libraries.
4. To restrict Ask AI to a pilot group before a wider rollout, go to the app’s Audience settings in Admin → Apps → All Apps and assign the group. Only members of that group will see the Ask AI entry point; others are redirected.
5. Confirm it works: open Apps → Ask AI → Chat and type a question. The response should stream in. If the organisation has other apps enabled (Shifts, Leave Management, Service Desk), try a question in each domain to verify agent routing.

3. How it fits together

Agents and routing — Ask AI is a dispatcher, not a single model. Each enabled app registers its own specialist agent. When a message arrives, Ask AI classifies the intent and routes it to the matching agent. With Cross-app reasoning & conversation continuity on, an agent can reach into another app to finish a request, so the employee never restates the question. Agents follow the same licence and audience rules as their parent app. If an employee cannot reach the Compensation app, the Compensation agent will not answer them either.

Company knowledge — Beyond each agent’s live data, Ask AI searches your organisation’s written knowledge. Each source is a separate switch under Company knowledge sources on the AI Features tab: Knowledge Base articles, HR Policies, SOPs, Site pages, Wiki pages, shared Drive files, Library items, Contracts, Posts, Verified answers, Training courses, Forms, Help articles, Recorder recordings, AI Notepad notes, custom app records, and the person’s own tasks. Every source applies the asking person’s own access. Within the Knowledge Base source, you can also switch off individual areas.

Libraries as a source — Answers use items from only the libraries a person can already open, and only when the Libraries app is open to them. A library restricted to people leaders reaches a leader’s answers and never an employee’s. An item that links to another record, such as a form or a wiki page, is cited only when the person can open that record too.

Knowledge scope for each surface — An embedded assistant (Ask AI shown on your intranet or another site, set up under Embedded assistants) can be narrowed to fewer sources. Its What it draws on list only narrows. It can never add a source your organisation has switched off, and each person’s access still applies. Leaving every box clear means “use everything the organisation allows”.

For example, a perks panel on the intranet home might tick only Library items. Asked “what discounts do we get?”, it answers from the perks library. Asked about the leave policy, it will not cite HR Policies, although the leave agent can still show the person their own balance.

Verified Knowledge — When Ask AI gives a wrong or incomplete answer in a News Feed thread, a user can submit a correction. These corrections land in the Verified Knowledge queue (pending review). An admin approves, edits, or rejects each one. Approved entries become part of the retrieval corpus and directly influence future answers.

Personal Memory — The assistant can remember facts about each employee across conversations. Every user can review and erase these facts from the Personal Memory tab. Whether facts are remembered is set in the Muscle Memory app, not in Ask AI.

Journeys — Multi-step goals the assistant tracks for an individual employee. An employee can have up to 3 active journeys at a time, each with up to 12 steps. The assistant updates progress during conversation, and the employee can override any step from the My Journeys tab. Journeys are switched on and off by Journeys (multi-step goal tracking) on the AI Features tab.

4. Running it

Reviewing Agent Guidelines

Agent Guidelines shape how every agent thinks, communicates, and acts. Go to Apps → Ask AI → Admin → Agent Guidelines. The tabs are:

1. Assistant — the assistant’s Name, Icon, Greeting and Description. The name is how it introduces itself in every answer.
2. Context & Rules — business context, custom instructions, universal guardrails, and a glossary, added to every agent’s instructions.
3. Communication — tone, response style, apology, and escalation messages.
4. Autonomous Actions — action-level guardrails, notification rules, and whether reasoning is shown in chat.
5. Overrides — per-agent customisations that override the global settings.
6. AI Features — organisation-wide switches, including the company knowledge sources.
7. Embedded assistants — Ask AI panels published on other sites, each with its own permissions and What it draws on list.
8. Questions — what people asked that Ask AI could not answer.

Context, Communication, and Autonomous Actions use a draft / publish workflow: changes stay in draft until you click Publish, then apply to all agents. Assistant, Overrides and AI Features apply on save.

Managing Verified Knowledge

Go to Apps → Ask AI → Admin → Verified Knowledge. The queue shows three tabs: Pending, Active, and Retired. Pending entries are submitted corrections awaiting review. Active entries are live and influence answers. Retired entries are archived.

For each pending entry, you can approve (optionally editing the answer), reject, or leave it. An approved entry immediately becomes part of the retrieval corpus.

Using the Knowledge Map

Apps → Ask AI → Admin → Knowledge Map shows the graph of content linked to skills and roles. This makes retrieval skill-aware: an answer about welding safety is prioritised for employees whose role requires welding skills. You can add manual links between content (Knowledge Base articles, HR Policies, SOPs) and skills, or filter the graph by skill or role type.

Running the Test Suite

Apps → Ask AI → Admin → Test Suite (accessible to admins and managers) lets you send example questions to the live agent pipeline and see which agent handled them, the response, and the routing match. Tests run against the published Agent Guidelines configuration.

Placing the Ask the assistant card on dashboards

The Ask the assistant dashboard card lets people ask a question from the home dashboard. It shows the assistant’s name, icon and greeting, a question box, and up to 4 suggested questions for the person’s role. A named assistant’s card reads “Ask” plus its name.

The card is never added to boards automatically. It leads the People Leaders and Employees starter boards, and an admin can place it on any other board. It shows only to people who can open Ask AI.

Admins also see a Top questions tile for the last 7 days: how many questions were asked, how many were answered, and the most common questions. A question appears only when at least 5 different people asked it, so no one can be identified from it.

Processing Data Requests

Data requests (GDPR Articles 15/17/20) are handled for every app at once, Ask AI included, at Admin → Retention & Legal Holds → Requests about a person. Find the person, check what each app holds, then choose Export a copy or Erase their data. Erasure is irreversible. Only admins (not app admins) can access this surface.

Reviewing Analytics

Apps → Ask AI → Admin → Analytics shows this tenant’s Ask AI usage: adoption trends, volume, per-user and per-day breakdowns, with CSV export.

Managing the Prompt Library

Apps → Ask AI → Prompt Library shows a gallery of reusable prompt templates. The gallery combines system-wide curated prompts with the tenant’s own org-private prompts. Any employee can browse and click Use in Ask AI to deep-link into the chat with the prompt prefilled. Admins can create, edit, and delete tenant-private prompts. Prompts are organised into categories: Productivity, Communication, Workflow, Analytics, and Operations.

Asking about an attached video

Open Attach files, select an MP4, MOV, or WebM clip up to two minutes and 100 MB, and wait for Video ready before sending a question. The mobile web chat links to the same responsive composer. Processing uses timestamped speech plus up to six sampled frames, so answers can point to evidence while explaining unreadable details and gaps between samples. A silent video uses visual samples. Retry interrupted processing or remove the clip to cancel. Private clip evidence expires after 24 hours; existing conversation answers follow the organization’s conversation retention.

5. Settings

Settings are split across three places: the Settings page, the AI Features tab, and each embedded assistant’s own form.

Settings page

Setting	Default	What it changes
Enable General Questions	On	Whether Ask AI answers open-ended questions beyond schedule and app-specific queries
Enable Schedule Queries	On	Whether Ask AI answers shift and schedule questions
Enable Action Execution	On	Whether agents can use their tools — look up live data and carry out actions (changes ask for the employee’s confirmation first). Off stops live lookups too; answers from your company’s knowledge continue
Enable Voice Mode	On	Whether employees can use voice conversations with Ask AI
Voice Daily Limit	30 minutes per user	Per-user daily cap on voice usage; 0 means unlimited
AI Voice	Marin	The voice used for spoken responses
Voice Response Style	Concise	How verbose voice answers are
Voice Speech Pace	Normal	Speed of spoken responses
Include Platform Help Articles in Answers	Off	Whether the assistant’s answers draw on platform help-centre articles in addition to company knowledge
Keep conversation history for	24 months	How long transcripts are retained before permanent deletion by the nightly sweep
Allow users to clear their conversation	On	Whether employees see the “Clear conversation” and “Reset AI completely” controls

AI Features tab (Agent Guidelines → AI Features)

Setting	Default	What it changes
Multi-agent answers	On	Whether a question spanning several areas is sent to each relevant agent and the answers combined
Shared knowledge search	On	Whether agents search the company knowledge sources ticked below it
Company knowledge sources	Each source on, except Library items	Which sources shared search may use. Library items may be off for organisations that already had libraries when it was added
Leave out sources overdue for review	Off	Whether answers skip content that Content governance has flagged as overdue for review
App explanations from Help Center	On	Whether each app’s agent answers “how does this work?” from that app’s help articles
Journeys (multi-step goal tracking)	On	Whether the assistant tracks multi-step goals for employees
Auto-detect goals	On	Whether the assistant starts tracking a journey on its own when a goal is clear
Weekly improvement proposals	Off	Whether admins get a weekly Inbox list of suggestions drawn from failed or low-rated turns
Cross-app reasoning & conversation continuity	On	Whether an agent can reach into other apps to finish a request
Smarter routing (tool-signal boost)	On	Whether routing is more decisive when a question clearly matches an app
Multi-step planner	Off	Whether a “do X then Y” request becomes a plan with read-only steps run automatically
Re-plan attempts (planner)	1	How many times the planner may refine a low-quality plan

Each embedded assistant (Agent Guidelines → Embedded assistants)

Setting	Default	What it changes
What it draws on	All clear (everything the organisation allows)	Narrows this assistant’s company-knowledge answers to the ticked sources only

6. More help

• Ask AI FAQ — specific setup and operating questions, including limits, thresholds, and troubleshooting
• Ask AI — What It Can Do and How to Govern It — the platform-level governance article covering AI operations, moderation, and cross-app controls
• Ask AI — the assistant answers questions about Ask AI from these articles