Automating your work using customer intents
If you want to automate repetitive tasks, such as tagging tickets or responding to common questions, but do not want to spend a lot of time writing rules triggers, Gorgias offers a much more convenient way to simplify rules by using detected customer intents.
Each time a new customer message arrives, we take a guess at figuring out the customer's intent. Intents include:
- and many others
Intent detection supports the 16 most common languages (see below for list of languages), so even if the customer is writing over to you in another language, we can detect the keywords and trigger the rule!
Intents can be used to automate your ticket processing workflow. Here is how.
Common use cases
Intents are used as triggers within rules
Even if customer intents are automatically detected on all customer messages, they do not trigger automated actions by themselves. In fact, intents are just additional information embedded within rules and the best way to think about intents is to consider them as automated, optimized triggers in rules.
The two following screen captures show the kind of rules logic that can be simplified by intents:
Without using customer message intents (long, tedious, non-comprehensive)
Using customer message intents (easy and comprehensive)
Tagging tickets using customer intents
- In Settings, go to Rules and click Create new rule
- Select WHEN new message in ticket and/or ticket created as a trigger
- Click on THEN, and select an IF statement
- Select message -> intents -> contains one of -> "shipping/status"
- Click on following THEN, select Add tag as an action, and type "ORDER-STATUS"
- Click Save
- Don't forget to activate the rule!
This example shows a simple rule tagging tickets for "shipping/status", ''order/cancel' and "refund/request" intents.
Now, each time you receive a customer message asking about an update on a delivery or an unusual delay, that ticket will be automatically tagged!
You can then filter tickets based on customer intents by creating views.
Setting auto-replies based on customer intent
Some customer intents are suitable for automatic replies such as "shipping/status". For that, you can set up a rule that would apply a specific macro when an intent is found on a new customer message.
Since we generally don't want to auto-reply multiple times in a given ticket, we recommend to add a "auto-reply" tag when auto-replying to tickets using intents and use an additional filter to avoid avoid auto-replying twice.
List of customer intents
Asks for a discount or needs help applying one
Deprecated. Merged in exchange/request
Requests to exchange a received product or get a replacement
Requests a status update on an ongoing exchange
Positive feedback from a customer
Negative feedback from customer
Requests to cancel an order before it ships
Requests to edit an order before it ships
Received order contains damaged item or customer needs troubleshooting
Received order contains wrong/missing item
A message that does not need a reply. The ticket can probably be closed*
General question on product, brand, policies ...
The customer expresses gratitude. The ticket can probably be closed**
Question about a specific product (usage, properties...)
Requests help chosing a product (size, color, model...)
Deprecated. Merged in refund/request
Requests to proceed a refund
Requests a status update on an ongoing refund. Can overlap with return/status
Deprecated. Merged in return/request
Requests to return a received product
Requests a status update on a received product
Requests to change shipping information (address, name)
Notifies of an error with a package (lost, not in the mailbox...)
Asks information on shipping (countries, usual delays)
Asks a status update on an expected order
Asks about the availability of a product
Requests to cancel a subscription (auto-shipment, monthly order)
Requests to edit a subscription (frequency, content...)
*other/thanks intent: This intent is often found on messages where the customer is concluding a ticket with a kind word or a thank you message to the agent. Examples of messages that correspond to this intent are "Great - thanks so much! Appreciate the quick response and follow up!", "Thanks so much! Will do!", "Oh ok cool. I think we'll all end up ordering one! Have a great night!" , "". Finding an 'other/thanks' intent on a message often means the ticket can be closed.
**other/no_reply: This intent is often found on messages that do not need to be replied to. Messages that correspond to this intent are for example "Your Amazon order has just been delivered" or "Call ended. Duration: 57s" or "A customer just ordered a ...". Finding an 'other/no_reply' intent on a message often means the ticket can be closed. Social media tags also fall into this category.
Intent detection works in most commonly spoken languages, so it's a huge advantage over writing custom rules with keywords! The main languages supported are, but 16 languages are officially supported in total:
- Arabic (ar)
- Chinese (PRC) (zh)
- Chinese (Taiwan) (zh-tw)
- Dutch (nl)
- English (en)
- German (de)
- French (fr)
- Italian (it)
- Portuguese (pt)
- Spanish (es)
- Japanese (ja)
- Korean (ko)
- Russian (ru)
- Polish (pl)
- Thai (th)
- Turkish (tr)
FAQ: Common questions
- How does intent classification work?
- Gorgias intent classification system is based on state-of-the-art artificial intelligence algorithms. There is nothing for you to do, it's fully automated!
- The intent found is wrong, what should I do?
- Unfortunately, that can happen as language comprehension is hard. There are usually a lot of ways to say the same thing. You can combine message intents with simple logic for a better result.
- Can I use message intents along with usual rules logic (IF ... THEN)?
- Of course. Please do so.
- Can a single message have multiple intents?
- Yes, each message can have multiple intents. Besides, if you use intents to auto tag tickets, tickets will be tagged with the intents of all messages it includes.
- Do all messages have an intent?
- No, there is a predefined list of possible intents. Moreover, when the system is not confident about a prediction, it will not detect the intent.
- Can I add an intent to the list of intents that can be detected?
- No, but you can definitely contact us any particular request of feedback you may have!