[No QA] [HOLD] Create concierge ai hub and docs

docs/_data/_routes.yml

@@ -103,6 +103,22 @@ platforms:
         title: Chat
         icon: /assets/images/chat-bubble.svg
         description: Use Expensify's chat feature to split bills, chat with employees, and manage payments.
+
+      - href: concierge-ai
+        title: Concierge AI
+        icon: /assets/images/concierge-avatar.svg
+        description: Learn about Concierge AI and what it can do.
+        articles:
+          - href: Concierge-Basics
+            title: Concierge Basics
+          - href: Concierge-Intelligence
+            title: Concierge Intelligence
+          - href: Expense-Assistant
+            title: Expense Assistant
+          - href: Support-Agent
+            title: Support Agent
+          - href: Assistant-Manager
+            title: Assistant Manager
 
       - href: workspaces
         title: Workspaces

docs/articles/new-expensify/concierge-ai/Assistant-Manager.md

@@ -0,0 +1,84 @@
+---
+title: Assistant Manager
+description: Let Concierge act as your Assistant Manager—automating report tasks like submitting, approving, reimbursing, and exporting.
+keywords: Concierge Assistant Manager, submit report Concierge, approve report, export report, report automation, Concierge chat, AI report support
+---
+<div id="new-expensify" markdown="1">
+
+Concierge can act as your Assistant Manager—automatically submitting, approving, reimbursing, and exporting reports on your behalf. Just send a message, and Concierge will take care of it.
+
+# What is the Concierge Assistant Manager?
+
+The Concierge Assistant Manager is your digital teammate for reports. It helps move reports through your workflow—whether you're submitting your own, approving others', or exporting to accounting software.
+
+# What report actions can the Concierge Assistant Manager handle?
+
+Concierge can assist with most common report tasks:
+
+- Submitting your own reports  
+- Approving reports submitted by your team  
+- Reimbursing approved reports (if reimbursement is enabled)  
+- Exporting reports to connected accounting software  
+- *Coming soon:* Setting up approval workflows  
+
+Whether you're finalizing your own expenses or clearing out team approvals, Concierge can help you get it done faster.
+
+# What should I say to the Concierge Assistant Manager?
+
+Use plain, natural language—no special commands needed. Here are a few examples:
+
+- “Submit my April travel report”  
+- “Approve Alice’s report for $312”  
+- “Reimburse the report with the Uber and lunch expenses”  
+- “Export Bob’s latest report to QuickBooks”
+
+Concierge will confirm the action and carry it out, or ask for more detail if needed.
+
+# How does the Concierge Assistant Manager know which report I mean?
+
+Concierge uses context from where you're chatting to figure it out:
+
+- **In a report chat:** It assumes you’re referring to that specific report.  
+- **In a DM with Concierge:** It may ask for more details (like report name, date, or amount).  
+- **In a workspace chat:** It prioritizes reports linked to that workspace.
+
+Say what feels natural—Concierge will do its best to match your intent to the correct report.
+
+# Can I approve or export multiple reports at once with the Assistant Manager?
+
+Yes. Just be clear in your message. Examples:
+
+- “Approve all pending reports from my team this week”  
+- “Export all reimbursed reports from October to NetSuite”
+
+If Concierge needs clarification, it will ask before taking action.
+
+# What if I make a mistake?
+
+No problem—just let Concierge know. If your role and workspace settings allow, it can often undo or adjust the action. If not, the issue will be escalated to a human for help.
+
+# Who can use the Concierge Assistant Manager?
+
+Anyone with the right permissions:
+
+- **Submitting reports:** All members  
+- **Approving reports:** Workspace Admins and Approvers  
+- **Reimbursing and exporting:** Domain Admins or connected accounting roles  
+
+If you try to take an action that requires a higher permission level, Concierge will let you know who can help.
+
+# FAQ
+
+## Can I use the Concierge Assistant Manager on mobile?
+
+Yes! On mobile, Tap Create at the bottom of your screen, select Start Chat, and type Concierge into the composer box to start chatting. 
+
+## Is there a way to see what Concierge did?
+
+Yes. Concierge will always reply with a confirmation message or follow-up. You can review the conversation in your Concierge chat history.
+
+## Does the Assistant Manager support bulk reimbursements?
+
+Yes, if you have reimbursement permissions and bulk actions are supported in your workspace setup, you can ask Concierge to reimburse multiple reports at once.
+
+</div>

docs/articles/new-expensify/concierge-ai/Concierge-Basics.md

@@ -0,0 +1,105 @@
+---
+title: Concierge Basics
+description: Get to know Concierge - your built-in support agent, expense assistant, and workflow manager.
+keywords: Concierge AI, support chat, contact Concierge, submit expenses, approve reports, AI help, human support, disable escalation
+---
+<div id="new-expensify" markdown="1">
+
+Concierge is your built-in Expensify assistant—here to answer questions, create expenses, manage reports, and speed up your workflows. It’s part AI, part human, and always available.
+
+# What is Concierge AI in New Expensify?
+
+Concierge is an intelligent support agent that lives inside Expensify. It’s powered by AI and backed by real humans when needed. Whether you’re asking about a feature, creating an expense, or troubleshooting an issue, Concierge is designed to help.
+
+Think of Concierge as your:
+- Product expert  
+- Expense assistant  
+- Workflow manager  
+
+All rolled into one!
+
+# Where to find Concierge AI in New Expensify
+
+You can chat with Concierge from anywhere in the app:
+
+- **On web:** Press the + button at the bottom left corner of the screen, select Start Chat, and type Concierge into the composer box.
+- **On mobile:** Tap Create at the bottom of your screen, select Start Chat, and type Concierge into the composer box.
+
+You can also contact Concierge by:
+- Email: **concierge@expensify.com**  
+- Text: **47777**
+
+# How to use Concierge AI
+
+Just start chatting—no special commands required. Ask a question, give an instruction, or upload a receipt.
+
+Here are a few examples you can type:
+- “How do I change my Workspace settings?”  
+- “Create a $12 lunch expense for today.”  
+- “Submit my April travel report.”  
+- “Approve Alice’s report with the Uber expense.”
+
+> **Tip:** The more specific you are, the better the response.
+
+# What can Concierge AI do?
+
+Concierge can help you:
+- Answer questions about Expensify’s features and settings  
+- Diagnose and troubleshoot issues  
+- Create or edit expenses  
+- Submit, approve, reimburse, or export reports  
+- Schedule a call with a human (when available)
+
+New features are added regularly.
+
+# Can I ask Concierge AI multiple questions at once?
+
+Yes! Just be clear in your message. Concierge can understand multi-step instructions like:
+
+> “Is this report reimbursable? And can you create a $25 taxi expense for it?”
+
+# Can I talk to a real person instead of Concierge AI?
+
+Absolutely. Just say you’d like to chat with a human, and we’ll connect you—24/7.
+
+If you’re working with an onboarding specialist or account manager, Concierge can schedule a call with them for you.
+
+# How to disable human escalation in Concierge AI
+
+Prefer an AI-only experience? You can prevent Concierge from escalating your messages to a human.
+
+**To disable escalation:**
+1. Go to `Account > Settings > Preferences`.  
+2. Toggle off **Allow Concierge to escalate to humans**.
+
+# FAQ
+
+## Is Concierge a real person?
+
+Not exactly. Concierge is an AI-powered assistant that helps with support, expenses, and approvals. But if it can’t help—or if you prefer a human—it’ll bring in a real person automatically.
+
+## Do I have to use special commands to talk to Concierge?
+
+Nope! Just type naturally, like you’re messaging a coworker. You can ask questions, give instructions, or even send receipts directly in the chat.
+
+## What happens if Concierge doesn’t understand my question?
+
+If something’s unclear, Concierge will ask for more details. And if it still can’t help, it’ll escalate to a human who can jump in.
+
+## Can I use Concierge on both web and mobile?
+
+Yes! Concierge is available everywhere—desktop, mobile app, email, and even SMS. Wherever you start a chat, Concierge will follow the conversation across platforms.
+
+## Is it safe to send receipts or sensitive info to Concierge?
+
+Yes. Concierge only sees what it needs to help with your request. Expensify doesn’t use your data to train public AI models, and we have zero-retention agreements with our partners.
+
+## What if I change my mind and want to talk to a human?
+
+Just say so! You can type “Talk to a human” at any time and you’ll be connected with a real support rep.
+
+## Can I turn off human support completely?
+
+Yes. If you’d rather stick with AI-only replies, you can disable human escalation from your preferences. You can do this via `Account > Settings > Preferences` and toggle off **Allow Concierge to escalate to humans**.
+
+</div>

docs/articles/new-expensify/concierge-ai/Concierge-Intelligence.md

@@ -0,0 +1,132 @@
+---
+title: Concierge Intelligence
+description: Learn how Concierge uses context, AI, and automation to understand what you need and get it done fast.
+keywords: Concierge Intelligence, how Concierge works, AI context, multi-modal agent, contextual chatbot, Expensify support AI, Concierge automation
+---
+<div id="new-expensify" markdown="1">
+
+Concierge isn’t just smart—it’s **context-aware**, multi-skilled, and built to help you get more done with less effort. This guide explains how Concierge works, what makes it intelligent, and how to get the most out of it.
+
+# How Concierge works
+
+Concierge is a **hybrid AI agent** built with GenAI, automation tools, and a powerful rules engine. It figures out what you're trying to do, pulls in the right tools, and responds with answers—or takes action automatically.
+
+If Concierge doesn’t know something, it escalates to a real person.
+
+# Key traits of Concierge intelligence
+
+Concierge uses a few smart capabilities to figure out how to help:
+
+- **Context-aware** – Understands *where* you’re asking from to tailor the response.
+- **Multi-modal** – Acts like multiple assistants (support agent, expense assistant, etc.) in one.
+- **Hybrid support** – Combines AI and humans, switching seamlessly when needed.
+- **Natural language understanding** – Responds to conversational input like a coworker would.
+
+# Understanding Concierge's contextual behavior
+
+Concierge adapts based on where and how you ask questions:
+
+- If you chat inside a **report** → It answers about that report.  
+- If you're in a **DM** → It considers your full expense history.  
+- If you're in a **Workspace chat** → It focuses on that Workspace.  
+- If you're on a specific **expense** → It assumes you're referring to that expense.
+
+You don’t need to explain every detail—just speak naturally and Concierge will fill in the blanks.
+
+# What is a multi-modal agent?
+
+Instead of having different bots for different tasks, Concierge is all-in-one. That means:
+
+- You can ask anything—no need to figure out who to ask.  
+- You can combine requests in one message.  
+  - Example: “Is this reimbursable? Can you add a $12 taxi expense too?”
+
+# Can I ask multiple things at once?
+
+Yes! Concierge can understand complex or multi-part requests, as long as it’s clear what you want.
+
+Here’s what works well:
+- “Create a $5 lunch expense and add it to my April report.”  
+- “What does ‘non-reimbursable’ mean, and can you mark this expense as such?”
+
+The more specific you are, the better Concierge can help.
+
+# What if Concierge doesn’t know something?
+
+If Concierge can’t answer a question, it will escalate to a real person—often without you needing to ask.
+
+You can also type **“Talk to a human”** at any time to request an escalation.
+
+# What powers Concierge behind the scenes?
+
+Concierge is powered by a mix of technology and human backup:
+
+- A custom-trained GenAI model  
+- A rules engine that handles logic and automation  
+- Real-time access to your Expensify data (just enough to help)  
+- Escalation paths to live support when needed
+
+# How does Concierge AI protect my data?
+
+Concierge is built with privacy at its core:
+- Your data stays inside Expensify and is only accessed as needed.  
+- AI systems never see more than what's required to respond to your message.  
+- We have **zero-retention** agreements with AI providers, meaning your data is never stored or reused.
+
+There’s no risk of another customer seeing your data—because they can’t.
+
+# Why Concierge is different
+
+- Understands where you’re chatting from and adjusts accordingly  
+- Handles natural, multi-part questions with ease  
+- Manages support, expenses, and reporting in one place  
+- Escalates to a human when needed  
+- No setup or training required—it just works  
+
+# FAQ
+
+## How does Concierge determine what I’m referring to?
+
+Concierge uses the context of your message—like the chat location (report, Workspace, expense), prior messages, and the specific wording of your request. It pairs this with internal identifiers (like report IDs or expense metadata) to match your request to the right object, without you needing to specify it directly.
+
+## How is context "understood" technically?
+
+Context is inferred using a combination of:
+- Chat metadata (where you're messaging from)
+- Your role and permissions
+- Previous conversation turns
+- Structured data (like expense amounts, merchant names, and report statuses)
+
+This context is then passed into the AI model along with your input to produce a tailored response.
+
+## What’s the difference between the AI and the rules engine?
+
+- The **AI (LLM)** interprets what you’re asking and generates a human-like response.
+- The **rules engine** executes structured logic behind the scenes—for example, determining which reports are ready to submit, or applying Workspace rules to an expense.
+- The two work together: AI figures out intent; the rules engine ensures valid outcomes.
+
+## What does "multi-modal" mean in this context?
+
+In Expensify, "multi-modal" means Concierge can handle multiple functions—support, expense management, and approvals—in a single thread. You don’t need to choose a specific agent or mode. Concierge identifies your intent based on the language used and responds accordingly, even if you blend requests.
+
+## How does Concierge know when to escalate to a human?
+
+If the model is unsure, detects missing context, or encounters something outside its capabilities (e.g., Workspace reconfiguration), it uses fallback logic to escalate. You can also manually escalate by saying “Talk to a human.”
+
+## Is Concierge really reading my data?
+
+Not exactly. The AI model doesn’t browse your data freely. It only receives the specific structured data relevant to your request—for example, the report name, amount, or receipt image you reference. This is passed securely via the system prompt, and the model can’t access or recall any other customer’s data.
+
+## Is my data ever stored or used to train Concierge?
+
+No. Concierge runs on AI models with **zero-retention agreements**. That means your data is not logged, stored, or used to train future versions of the model—by Expensify or by any third party.
+
+## Can Concierge give different answers in different places?
+
+Yes—and that’s by design. Because it’s context-aware, asking “Is this reimbursable?” in a report chat will return a different result than asking it in a DM. This makes responses faster and more relevant.
+
+## Can I test how Concierge behaves in different modes?
+
+Absolutely. Try sending the same question in different contexts (a report, a Workspace chat, a DM) and you’ll see how Concierge tailors its response to match.
+
+</div>