Recent searches


No recent searches

Dakota's Avatar

Dakota

Joined Apr 15, 2021

·

Last activity Jan 23, 2024

Zendesk Luminary

Following

0

Followers

0

Total activity

24

Votes

2

Subscriptions

13

ACTIVITY OVERVIEW

Latest activity by Dakota

Dakota commented,

Community comment Discussion - Artificial Intelligence

Hi everyone,

Thanks for sharing how your teams are leveraging AI! It's really great to see the creative ways companies are implementing this technology.

At our company, we've been continuing to enhance our AI journey and are excited about new possibilities. So far, we have started using AI for:

Analytics: We are feeding historical chat transcripts into an AI analytics platform to uncover insights about common customer questions, pain points, and suggestions. This is helping guide our self-service content strategy.

Ticket Summarization: For especially long or complex tickets, we are testing an AI summary tools from Tethr that digests the key details into a concise summary. This helps agents process the details more efficiently.

Intent Detection: We have started to use Zendesk AI to analyze visitor questions and predict their intent. Based on the intent, we've created many autoreply bots that can provide tailored responses to common inquiries.

Future Initiatives: Our team is exploring ways to enhance our chatbot capabilities further using generative AI. We are also utilizing and enhancing our AI-powered quality assurance review tool to analyze agent tones and messages.

As we continue our AI adoption, we hope to learn from the community's experiences here and love to see everyone's comments.

View comment · Posted Jan 23, 2024 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - AI and automation

Hi Victoria,

I would highly recommend looking into Claude AI, I've had great success using this for surface-level analytics of data like user reviews and interactions with names and private info removed. However, its training is strictly based on public domain information in order to provide responses.

I've found that Claude AI works much better than Chat GPT for more personable responses and can ingest a much larger dataset than GPT (I believe up to 50k words). Which, if you were trying to leverage more internal info that doesn't contain sensitive info, you can copy and paste that internal article or info into Claude, and prompt the AI to analyze or respond piggybacking what you pasted.

You might also want to explore their customized enterprise AI tools so that your own teams have access to vetted and trained info on your own knowledge bases in a controlled, secure environment. That way they could leverage your private content to better assist employees and customers, with proper data permissions and privacy protections as assessed by your internal IT. 

Feel free to send me a message if you want more info about my experiences!

View comment · Posted Dec 07, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - Tickets and email

Hi Evan,

It sounds like you have ticket forms enabled that require a ticket description field. However, that field is not showing for you as a full agent when trying to create an organization ticket.

The key difference is that light agents have access to an optional ticket description field on forms, while full agents do not, as full agents have access to the full ticket editor instead.

There are a couple of options I'd recommend exploring:

  1. Edit the form used for organization tickets and make the description field optional rather than required. This would prevent the error. Full agents would still not see the field but could submit successfully.
  2. For organization tickets where you need a description from full agents too, I'd suggest creating a custom text field called "Description" on the form instead. This would allow both light and full agents to populate that field when creating org tickets via forms.

Hopefully, this will help steer you in the right direction!

View comment · Posted Dec 06, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - Tickets and email

Yes it is possible to pass data from one form to another in Zendesk. :) Here are a few options to accomplish this:

  1. Using Tags:
  • When setting up Form 1, create a text/numeric tag for each value you want to pass (ex: "customer_name", "ticket_priority").
  • Use Post-submission triggers on Form 1 to set those custom tags with submitted values.
  • In Form 2, reference the tags using merge fields like {{ticket.custom_fields.customer_name}}
  1. URL Parameters:
  • On Form 1's redirect URL, append parameters like ?customerName={{ticket.requester.name}}
  • In Form 2, reference the parameters like {{ticket.custom_fields.url_parameter_customerName}}
  1. Update with Triggers:
  • Create a Post-submission trigger on Form 1 that selects Form 2 ticket and uses the Zendesk API endpoint /api/v2/tickets/{id}.json to update ticket fields with submitted values
  • Fields will now be populated with the trigger-updated values
  1. Server Side Script:
  • Implement a workflow using a tool like Zapier that connects your form to a serverless function
  • Write the function to insert submitted Form 1 values into defined fields in Form 2 using the Zendesk API
  • This keeps the data sync entirely on the backend

The key for all options is making use of those custom tags, URL parameters, and API endpoints to dynamically link data between two separate form tickets/requests.

View comment · Posted Dec 06, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - Chat, messaging, and widgets

Here are a few options to help filter out the bot messages when analyzing agent messages in the Messaging Tickets dataset:

  1. Use the "Assignee ID" field to differentiate bot agents vs human agents. You can create a metric like "Assigned Human Agent Messages" filtered to tickets where:
Assignee Type = 'Agent'
AND
Assignee ID !=  
AND 
Assignee ID != 

This will only count messages by human agents.

  1. Create a custom tag or field at the ticket level to denote "Bot Assisted" tickets. Then filter these out.
  2. Use the Ticket Audits dataset and look at changes to the "Assignee ID" field. Where it flips to a bot and back to a human, you can selectively count messages before and after.
  3. Use Zendesk API to pull all messages and programmatically filter out ones from known bot IDs based on the "author_id" in the message.

Hope this helps!

View comment · Posted Dec 01, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - Chat, messaging, and widgets

Here are a few options to better style and customize the Zendesk chat widget:

  1. Use CSS overrides:

You can provide custom CSS rules to modify colors, sizes, fonts etc. Documentation here:

https://develop.zendesk.com/hc/en-us/articles/360001948687-App-framework-CSS-overrides

For example, to change background color:

.zendesk-web-widget-root {
  background-color: #2196F3 !important;
}
  1. Enable widget theming:

In the account settings, administrators can enable "theming" which allows changing colors and logo. This provides configuration options instead of custom CSS.

  1. Use the Embeddable SDK:

This allows providing your own UI wrapped around the Zendesk Chat SDK. Fully customizable look and feel. More advanced integration.

Docs: https://developer.zendesk.com/embeddables/docs/chat-sdk/introduction

 

View comment · Posted Dec 01, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community comment Q&A - AI and automation

I'm no coding wizard but I believe it would require some level of backend coding.

Would likely require something like this. When the user logs into your website, store their name (e.g. "Josh") in a session variable or some other server-side storage that you can access later.
In your chatbot code, retrieve the logged-in username from the server-side storage before configuring the chatbot.

For example:

// Get logged in user's name
const name = getLoggedInUserName(); 

// Configure chatbot 
const chatbot = new Chatbot();

// Customize greeting
chatbot.setGreeting(`Hi ${name}, I'm here to help.`);
The chatbot should now greet the user with their name fetched from the server-side storage:
"Hi Josh, I'm here to help."

View comment · Posted Dec 01, 2023 · Dakota

0

Followers

0

Votes

0

Comments


Dakota commented,

Community commentZendesk Explore EAP - Quick reports using natural language processing

Thanks for providing the opportunity to share our feedback, here are some questions I'd be interested in:

  • What are the top 5 requested features or enhancements from customer support tickets over the past month?
  • What are the top 10 praises and pain points related to x issue/topic over the last week?
  • Which agents closed the most tickets last month and how does that compare to their averages
  • What are the top negative phrases from customer support tickets over the past day/week/month/quarter?
  • What percentage of chat/email conversations have been classified as frustrated in the past week?
  • Compare this week's average first response time to the same week last year.
  • Which 3 knowledge-base articles had the most views this month?
  • Provide me with a list of common phrases or topics that are not covered with knowledge-based articles.
  • How many tickets were created per day this week versus the weekly average for the past 6 months?
  • Which topic areas have seen the biggest/least improvement in CSAT score over the last 6 months?

Edit:
Adding more that I thought of.

  • Show me the top 3 reasons for cancellations this month that we haven't seen in previous months.
  • What is the predicted ticket volume for next Monday based on historical trends?
  • Which article topics have seen increased traffic over the past week that are not typically popular?
  • Build a graph showing the correlation between first reply time and customer satisfaction rating.

View comment · Edited Dec 01, 2023 · Dakota

0

Followers

1

Vote

0

Comments


Dakota created a post,

Post Feedback - Ticketing system (Support)

Is there any way to add a prompt like "Are you sure you want to join a chat?" when you click the 'In Queue' button just to prevent accidentally being forced into a live chat?

Twice today I have accidentally clicked that 'In Queue' button instead of the search button and ultimately had to respond to customers before transferring them to actual agents.

This would be so awesome to have, thank you!

Posted Mar 23, 2021 · Dakota

7

Followers

3

Votes

0

Comments