AI Agent explainedUpdated 5 hours ago
Gorgias’s AI Agent helps Shopify brands to deliver personalized customer experiences by instantly resolving repetitive inquiries and driving sales through AI-powered conversations.
AI Agent is trained specifically on your brand’s policies, website, and Shopify data to autonomously resolve customer questions — whether they’re visiting your store as a first-time shopper or looking for help with an existing order.
With its two skillsets — Shopping Assistant and Support Agent — AI Agent seamlessly handles pre-sales conversations, including recommendations, upsells and discounts, along with post-sales order tracking, edits, returns and subscription management.
From your Gorgias helpdesk, you can Train, Test, Deploy, Engage and Optimize your AI Agent with control and flexibility.
Below you’ll learn about AI Agent’s features and capabilities.
Requirements
- You must have an AI Agent subscription
- You must have a Shopify store connected to Gorgias
- AI Agent is not supported on BigCommerce, Magento or WooCommerce stores
Train
Knowledge sources
From your Gorgias helpdesk, you can control, maintain and update the knowledge that AI Agent uses to accurately respond to requests from customers. AI Agent learns from multiple sources, including your Shopify customer and order data, your website, Help Center articles, public URLs and external documents.
Tone of voice
Following your instructions, AI Agent maintains a consistent tone of voice — whether that’s professional, friendly, sophisticated or something else. This ensures every message reflects the experience that customers expect from your brand.
Guidance
Guidance lets you give AI Agent custom instructions about how it should handle specific topics and questions. Use Guidance to make sure the AI always follows your support policies, sales playbook and preferences when engaging with customers.
Actions
Create Actions that allow AI Agent to automate more complex tasks, and fully resolve tickets without human intervention. Actions connect AI Agent to your other business tools so that it can perform tasks across different apps, like cancelling an order, processing a return or modifying a subscription.
Multi-language resolution
AI Agent writes and responds to customers in any language supported by our large language models (LLMs) — over 80+ languages. You can control which language AI Agent uses to reply to customers in its language settings.
AI Agent can use any of your knowledge sources when responding in another language, even if the content has not been localized to the language the AI replies in.
Vision
AI Agent can read and understand images so that it can help customers who share image files in email tickets — whether it’s a photo of a damaged item, an order issue or proof of return. This means that AI Agent can successfully handle tickets with visual content that would otherwise have required a human to manually review, reducing the overall volume for your support agents.
Test
You can use real customer profiles and existing tickets to test how AI Agent responds to questions. Use the test environment to make sure your content and policies are up to date, and gain confidence with AI Agent before going live.
Audience testing
Test how AI Agent replies to different types of customers, on different channels. You can test responses for a new customer, an existing customer or an existing ticket on both email and chat to see how AI Agent responds. Evaluate how AI Agent responds to different scenarios and make tweaks to its knowledge and settings.
Answer reasoning
AI Agent gives you full visibility into how it responds to a question. See which knowledge sources AI Agent used to inform its response and what actions it took with the ticket (handover, snoozed, closed, and so on).
Deploy
AI Agent is always available across your sales and support channels. You can turn on AI Agent on Email and Chat, with more channels to come.
AI Agent on Email
AI Agent provides immediate, accurate responses to customer questions from emails while filtering out spam and other illegitimate inquiries. AI Agent pulls relevant information from your knowledge sources, the customer’s profile and order history to respond with personalized, on-brand answers.
If AI Agent cannot provide a response or detects frustration from the customer, it hands the ticket over to your human team. All responses go through a quality check using a separate AI model before replying to a customer.
AI Agent on Chat
AI Agent greets customers and responds with real-time answers, tailored to fit the experience of messaging on chat. On chat, AI Agent continues to use your tone of voice instructions to always reply on-brand, but adapts its messages to be quicker and more conversational than email.
When necessary, AI Agent hands over chat conversations to your team, or captures the customer’s email address to follow-up later.
Handover to humans
AI Agent automatically hands over conversations to your human team whenever it lacks confidence in an answer, encounters a listed handover topic, or does not find any relevant knowledge to answer a customer’s question. It also hands over if it detects that a shopper is angry or asks to speak with a human.
You can customize handover topics in your AI Agent settings. For chat conversations, you provide custom instructions for how AI Agent should hand over to your team in different scenarios, like when a chat is online or offline.
Engage
Proactive customer engagement [Requires upgrade]
With its shopping assistant skills , AI Agent acts as an always-on virtual assistant on chat and email, helping shoppers find and buy products on your website.
Shopping Assistant answers questions and recommends products to customers based on their behavior and browsing history. It offers discounts based on shoppers’ purchase intent when needed, and suggests relevant upsells and complementary items based on products the shopper has viewed, added to their cart, or asked about.
To encourage conversations with Shopping Assistant, you can turn on Customer Engagement features designed to meet shoppers where they are. Use settings like Suggested Product Questions, Ask Anything Input and Trigger on Search to create interest, guide product discovery and overcome hesitations.
Optimize
Direct feedback on tickets
You can coach AI Agent by giving it feedback on the tickets it answers. Here, you’ll see and rate the knowledge sources that AI Agent used in its response, and if necessary select a different piece of knowledge that AI Agent should have used instead.
Your rating and feedback directly influence how AI Agent approaches similar questions in the future.
AI reasoning [Coming Soon]
Get full visibility into AI Agent’s decision-making. With each response in a ticket, you can see why AI Agent chose to respond in the way it did. If a response doesn’t meet your expectations, you can edit or update knowledge directly from the view ticket to correct its behavior.
Performance reports
Your AI Agent performance reports show you where AI Agent is succeeding and where it might need to be optimized further. Track support metrics like automation rate, average CSAT, and first-response time. See what ticket topics AI Agent has been handling and their outcome.
For pre-sales conversations, you can see how AI Agent performed as a Shopping Assistant. Review metrics like influenced revenue, average order value and conversions to track AI Agent’s effectiveness at helping shoppers make a purchase.
Optimize page
You can use the Optimize page to explore topics that AI Agent has been handling. Here you can spot trends and find opportunities to improve AI Agent’s response, increase CSAT and expand its ticket coverage.
