Ask AI
Instant answers from your knowledge base.
MangoApps
Overview
Ask AI routes each question to whichever specialist agent owns the answer — scheduling for shift questions, leave management for PTO balances, service desk for ticket status, and more, one agent per MangoApps app — instead of a single model guessing from a document index. Because each agent is wired into that app's own live data, answers reflect what's actually in your schedule or ticket queue, not just what a policy document says.
When a request spans two areas, one agent hands the rest to another app's assistant mid-conversation, so a single question can cover both without being asked twice. Anything that deletes or removes data always pauses for your explicit confirmation, and most other actions that write data do the same.
Available from a web sidebar, the mobile app, and a browser extension, with one conversation carrying across all three.
Highlights
Capabilities
AI Chat & Conversation
-
Chat from a web sidebar, the mobile app, and a browser extension — one conversation synced across all three
-
Multi-turn follow-ups that keep context, including short/ambiguous ones ("two weeks ago", "no", "those")
-
Streaming, token-by-token responses — the answer appears as it's written
-
Rich, interactive answers — inline tables, charts, timelines, report cards, and clickable action buttons
-
@mention a specific agent inline to force routing (e.g. @hr, @scheduling)
-
Multi-lingual greetings and per-language response translation
-
In-app Prompt Library of reusable starter prompts (central + tenant-authored)
-
Searchable conversation history across every past question and answer
Proactivity, Memory & Journeys
-
The assistant prompts you — playbook results and approvals arrive in your Inbox instead of waiting to be asked
-
Reply to a proactive Inbox item and the assistant answers in-thread, grounded in the item
-
Cross-session personal memory of durable per-user facts, with a private review/erase page — hands off to the separate Muscle Memory app (its own license) once an admin enables it
-
Journeys — tracks a multi-step goal (onboarding, a migration) across sessions and guides you step by step, on a private My Journeys page
-
Auto-detect goals — optionally starts tracking a multi-step goal it notices, transparently and one-click dismissible (admin opt-in)
-
Weekly self-improvement — analyzes failed / misrouted / low-rated turns and sends admins concrete fixes
Agent Routing
-
Automatic routing of each question to the best-fit specialist agent
-
Sticky routing keeps a multi-turn session with one agent ("Continuing with: <Agent>")
-
Cross-app handoff — one agent hands the goal to another app's assistant
-
Plugin agent proxies for third-party marketplace extensions
-
Number of built-in domain agents: 80+
Knowledge & Grounding (RAG)
-
Grounds answers across your knowledge base, HR policies, SOPs, wikis, Sites pages, and shared Drive files
-
Admins choose which knowledge sources the AI is allowed to draw on
-
Optional platform help articles included in answers (admin toggle)
-
Semantic, meaning-based retrieval finds the right passage even when the wording differs
-
Every answer shows how it was produced — Computed, Sourced, or Verified badges with citations
-
Company Brain — a user's correction becomes admin-approved verified knowledge; News Feed "verified answers" flow in too
-
Live data pulled from connected apps via agent tools (schedules, leave, tickets)
-
Browser extension grounds answers in the LinkedIn, email, or calendar page you're viewing
Tool Calling & Actions
-
Agents call live, read-only tools for real-time data across apps
-
Write/risky actions require explicit user confirmation, with multi-step clarification before acting
-
See a before/after diff preview of any change before you approve it
-
Human handoff — escalate to a person, or auto-create a service desk ticket when confidence is low
-
Attach files to a question — image (PNG/JPG/WebP/GIF) vision + PDF text extraction (Word, Excel, and CSV files can also be attached)
Voice & Multimodal
-
Voice mode — speak a question, hear a spoken answer
-
Live, hands-free voice conversation you can speak into and interrupt naturally
-
Admin-configurable AI voice, response style (concise/balanced/detailed), and speech pace — one tenant-wide default, set with an in-settings preview
Safety, Audit & Admin Control
-
Content moderation with prompt-injection / system-prompt-security checks
-
Business-scoped tenant isolation — data never crosses organizations
-
Admin Agent Guidelines + per-agent instruction overrides
-
Admin toggles to disable action execution, voice, or help-article grounding
-
Admin review queue to approve, edit, or retire verified-knowledge (Company Brain) entries
Limits & Specs
-
Voice daily limit (admin-configurable): 30 minutes per user (0 = unlimited)
-
Voice session limit: 15 minutes per session
-
RAG token budget / context window: 128,000 tokens
-
Attachments per message: Up to 3 files, 8 MB each; PDFs up to 100 pages
-
Pricing: Included with MangoApps Workforce (requires license)
Screenshots
Use cases
FAQ
Responses are grounded in your organization's knowledge base, admin-verified corrections ("Company Brain"), any files you attach to a message, and live data pulled from connected apps through agent tools. Platform help articles can optionally be included by an admin.
From a chat sidebar in the web app, the mobile app, and a Chrome browser extension. All three route through the same agentic pipeline and share one conversation, so you can start a question on the web and continue it on mobile. In the browser extension, answers can also be grounded in the LinkedIn, email, or calendar page you're currently viewing.
Each question is automatically routed to the best-fit specialist agent out of 80+ built-in domain agents (scheduling, HR, leave, service desk, reporting, safety, and more). Multi-turn sessions stick with the same agent ("Continuing with: "), one agent can hand a goal off to another app's assistant, and you can force routing by @mentioning an agent inline (e.g. @hr, @scheduling).
Specialized agents can both read live data (schedules, leave balances, tickets, reports) and perform actions. Any action that writes or changes data always pauses for your explicit confirmation — with a before/after diff preview — and the AI never writes on its own.
Yes — up to 3 files per message, 8 MB each. Images (PNG, JPG, WebP, GIF) are analyzed with vision, and PDFs are read via text extraction (up to 100 pages; scanned image-only PDFs may have no extractable text). Word, Excel, and CSV files can be attached but are not currently parsed into text, so the AI won't read their contents.
Yes. Voice mode lets you speak a question and hear a spoken answer, including a live, hands-free OpenAI Realtime conversation you can interrupt naturally. Your admin sets one AI voice, response style, and speech pace for the whole organization, chosen with an in-settings preview. Voice is metered: sessions are capped at 15 minutes, plus an admin-configurable daily limit (30 minutes per user by default, or unlimited).
Ask AI is the unified conversational assistant that routes each question across every specialist agent and grounds answers in org-wide knowledge. Personal Agent is one of those specialists, hard-scoped to your own bookmarks, drafts, tasks, and inbox. Workforce Bots are standalone document-Q&A bots (HR Policy, SOP, Compliance, etc.) that admins install against specific knowledge bases. You reach the same intelligence through Ask AI's single chat surface.
Conversations are scoped to your business tenant and never shared across organizations, with content moderation and prompt-injection checks on every request. History is per-user — each user has their own searchable conversation, not a shared thread.
Journeys — the multi-step goals Ask AI tracks with you, like onboarding a hire or migrating a policy — are on by default, private to you, and managed on your My Journeys page; admins can turn Journeys or auto-detect off under Agent Guidelines → AI Features. Durable cross-session memory of personal facts (your preferences, what you're working on) is a separate capability provided by the Muscle Memory app: once an admin enables it and its "Remember durable facts for each user" setting, those facts are strictly per-user — managers and admins never see them — and you can review or erase them on your private My Memory page.
When confidence is low it says so rather than guessing, and can offer a handoff to a person; below a confidence threshold it can automatically create a service desk ticket so the request is tracked to resolution.
