Georgie

AI Customer Service

Customer Info

Ticket

#

Tags

Recent Orders

Enter a ticket ID and click Load to see customer info

Knowledge Base

Policies

Corrections

FAQ (from Gorgias Help Center)

Analytics

Tickets Processed

Drafts Generated

Escalations

Auto-Replies

Spam Rescued

Tokens Used

AI Cost

Adjust AI Cost

Add a cost offset for usage before tracking was enabled.

$

By Channel

By Intent

By Brand

AI Copilot Usage

Agent Messages

AI Responses

Knowledge Base

Corrections

FAQ Entries

Daily Volume

Users

Team Members

Change Password

Response History

responses

Response Detail — Ticket #

Brand

Intent

Channel

Action

Customer Message

AI Draft Response

Escalation Reason

AI Reasoning (System Prompt)
Knowledge Base Context Used
Order Context

Feedback

Saved automatically. This feedback becomes a correction in the Knowledge Base.

Saved feedback:

Add Feedback Note

Escalation Keywords

Global setting — applies to all brands. When any keyword appears in a customer message, the ticket is automatically escalated to a human agent.

Assign Escalated Tickets To

When a ticket is escalated, it will be automatically assigned to this agent in Gorgias.

How it works

  • Keywords are matched case-insensitively in customer messages
  • GPT-4o-mini also detects angry/frustrated sentiment — escalates even without keyword matches
  • Low confidence classifications (<50%) are auto-escalated as well
  • When triggered, the ticket is tagged "ai-escalated" and no AI draft is generated
  • Escalated tickets are assigned to the selected agent above (if configured)

Ticket Scan

tickets

Scanning recent tickets...

Fetching tickets from Gorgias, analyzing agent responses, generating recommendations. This takes 1-2 minutes.

Previous Scans

Backtest / Simulation

Run Backtest

Test the AI pipeline on historical resolved tickets. Georgie re-processes each ticket and compares its draft to what your team actually sent.

tickets (max 50)

Processing historical tickets through the AI pipeline...

Each ticket goes through classification, context building, draft generation, and QA. This takes 1-3 minutes.

How it works

  • Fetches recent resolved tickets from Gorgias where an agent already replied
  • Re-runs the full AI pipeline: classification, context building, draft generation, and QA validation
  • Compares the AI draft to the actual agent reply using word overlap similarity
  • Higher similarity = AI drafts are closer to what your team would send
  • Use this to test changes to prompts, policies, or KB entries before they go live

Follow-Ups

No follow-ups found.

Follow-ups are auto-scheduled for return (3 days) and WISMO (2 days) intents.

How it works

  • Follow-ups are auto-scheduled when Georgie processes return (3-day) or WISMO (2-day) tickets
  • A background task checks hourly for due follow-ups and posts reminder notes to Gorgias
  • You can manually complete or cancel follow-ups from this page

Auto-Closed Tickets

auto-closed (30d)

How it works

  • Georgie automatically detects emails from newsletter and automated senders (Alibaba, AliExpress, Reputon)
  • Matching tickets are tagged "auto-closed-newsletter" + "georgie-processed" and closed immediately — no AI draft is generated
  • Detection works by exact sender match (e.g. noreply@member.alibaba.com) and domain wildcard (e.g. any @aliexpress.com address)
  • Every auto-close is logged here with the sender, subject, and body preview for auditing
  • Use the search bar to find specific senders or subjects, and the brand dropdown to filter by brand

How It Works

Open in Notion Last synced:
Loading...
No wiki pages synced yet.

Social History

actions
Time Brand Platform Type Content Label Sentiment Action Reply Actions
Page of

Social Analytics

Comments Received

Comments Hidden

Comments Replied

DMs Replied

Smart Label Distribution

Daily Activity

Date Comments Hidden Replied Escalated Activity

Keyword Blacklist

Comments containing these keywords are auto-hidden

Add Keyword

Bulk Import

Active Keywords

Keyword Match Count Brand Added

AI Insights

Recommendations

Social Settings

Brand Settings

AI Model

Model

Input Price

Output Price

Session Cost

Total Tokens

API Requests

Input / Output

Add Correction

Rules, guidance, examples, and instructions for how the AI should handle tickets for this brand. Free-form text — write as much as you need.