Kustomer IQ

Learn more about how to predict customer's needs with Kustomer IQ.
Categories
  • Boost agent performance with Agent Assist

    Agent Assist is powered by several AI engines, including the ChatGPT engine from OpenAI, and uses the latest machine learning technology to supercharge your support team's workflow. Who can access this feature?User typesAdmins can turn on AI features for agents. Agents can access the features from the reply editor.To learn more about Agent Assist for Agents, visit our Kustomer University course.In this articlePricing detailsTurn on Agent Assist for your organizationAccess the AI featuresText enhancement optionsExpand textFix spelling and grammarConversation summariesManually summarize a conversationAutomatically summarize conversational assistant conversationsAutomatically generate insightful subject linesGenerate a responseTwo-way message translationPricing detailsYou can purchase Agent Assist for $39 per seat per month. If you exceed your monthly allotment of translated words included as part of the per license fee, you will be billed at the rate of $0.00050 per translated word above such allotment, billed at the end of each month. Please fill out this form for any questions related to pricing.Turn on Agent Assist for your organizationYou can trial Agent Assist for free for 14-days before upgrading to include these features in your plan. To do so, go to Settings > Kustomer IQ > Agent Assist. Note: Agent Assist uses experimental technology to create AI-generated content for your team. When using any of these features, review the results before sending them. None of this output is automatically sent out to users before you can review it. Your data is not being used to train or customize any machine-learning model. For more information, see Open AI's API data usage policies.Once you accept the terms of use, you can access each setting on the Agent Assist page. The Test Mode option will be turned on by default. While turned on, only users in your organization with admin permissions can access these features. This helps you evaluate these features before introducing them to the rest of your team.Note: Remember to turn the Test Mode setting off when you are ready to grant AI feature access to agents.The rest of this article will cover each AI-powered feature in further detail.Access the AI featuresWhen responding to a customer, you can use the AI Enhancements icon in the reply editor to access the expanded text, fix spelling and grammar, and generate response features. The options will dynamically appear in the editor as you type your response. You can review and edit all AI-generated suggestions before sending the message.Text enhancement optionsAgent Assist offers a variety of actions that leverage AI to optimize text.Expand textYou can type in a few words and use this option to generate a response with a more formal, helpful tone. Fix spelling and grammarUse this option to correct spelling and grammar in your reply. The changes are incorporated automatically without having to review them individually. Conversation summariesAgent Assist offers different settings that allow conversations to be summarized so agents can get quicker context.Manually summarize a conversationWhile composing a note, agents can use Summarize Conversation to add an AI-generated conversation summary to the note. This feature reviews all of the messages in the current conversation and adds a bulleted list to the note that outlines what was discussed. If an agent needs to hand a conversation off to another team member, this summary makes it easier for the next person to understand what's happening without reading through every message. Automatically summarize conversational assistant conversationsYou can enable conversational assistants to automatically summarize customer interactions and post notes, letting agents instantly gain context without reading through earlier messages.Summaries are only generated if a conversation is detected to have a significant amount of information in it. For example, a summary won't be created if a customer only gives yes/no answers to the assistant and it's quicker to read the original messages.Automatically generate insightful subject linesKustomer normally generates an automatic conversation subject using the first 6 words of an initial message for channels that don't have a dedicated subject line field, such as chat or SMS. This sometimes results in subjects with vague or unhelpful names, making it difficult to find a relevant conversation on the timeline.Turning on this setting will automatically generate a succinct subject so agents have insight into a conversation before they open it. Note: Subject lines will only be automatically generated for conversational assistants using the SMS or chat channel.Generate a responseThis feature analyzes conversations and lets you quickly compose messages using information from your organization's knowledge base.Note: To ensure better responses, we recommend having a knowledge base with comprehensive articles that contain information that can help answer commonly asked questions for your organization.When activating this feature on the Agent Assist settings page, you can select the knowledge base and article type you want to use from the drop-down menus that appear. Public, internal, or both public and internal articles can be used to form responses.Responses generated using content from a public knowledge base will include a direct link to the article.Two-way message translationThe two-way translation feature lets agents quickly translate both customer messages and their replies directly within a conversation. All organizations will receive 10,000 translated words* per seat per month to use across the entire organization. This means that instead of allocating this limit individually per seat, the limit is shared across all agents within your organization. For example, if your organization has 4 seats, you will have 40,000 translated words that can be used amongst your agents. An example monthly breakdown for this limit is:Agent 1 used 15,000 translated wordsAgent 2 used 10,000 translated wordsAgent 3 used 5,000 translated wordsAgent 4 used 10,000 translated wordsOrganizations that exceed their monthly limit will be billed $0.00050 per translated word over their limit. This monthly limit applies to both inbound and outbound messages.*What is a translated word?A translated word is equal to 5 characters, including spaces and punctuation. For example, “Return my order” counts as 3 translated words since it equals 15 characters. We will always round down to the nearest 5 when calculating translated word usage. For example, "Return status?" will count as 2 translated words since it equals 14 characters.Use the translation featureOnce the feature is turned on, agents can translate incoming messages to the application language set in their personal preferences.An agent can then reply to the customer and translate the response to the customer's preferred language before sending the response. The message will be translated into the default language set in the conversation's attributes.You can also change the language the message is translated to using the drop-down menu. Messages can be translated into the languages available on the Kustomer platform. The following languages are currently not supported:Belarusian (be)Kyrgyz (ky)Luxembourgish (lb)Montenegrin (sr_me)Norwegian (no)Twi (tw)Related articlesAgent Assist report
  • Automatic language detection

    Kustomer comes with a default Natural Language Processing (NLP) workflow that automatically detects the sentiment of an inbound email message. You can turn on an additional feature that automatically detects the language used in a conversation using Amazon Comprehend's dominant language detection. With this feature, the conversation's Default Language attribute is automatically set once the language used in the interaction is detected. This information can then be used to perform actions such as, routing all conversations of a particular language to the correct team or viewing all conversations for a specific language in a search.Who can access this feature?User typesAdmins can access these settings.In this articleTurn on automatic language detectionView a conversation's default languageTurn on automatic language detectionTo automatically detect which language your customers are using, you must turn the feature on via the Intelligence app.Go to Appsand select Intelligence.Select the Detect language automatically checkbox.Select Save Changes.With this setting on, Kustomer will now detect the languages used in these messages along with detecting the sentiment of incoming messages.View a conversation's default languageTo view the language your customer's are using, we recommend adding the Conversation Default Language attribute to the Conversation Insight Details card.Note: The Customer Default Language attribute does not impact the language that gets used in snippets. Go to Settings  > Platform > Klasses.Select Conversation and go to the Insight Cards tab.Select the Insight Details card.Search for the Conversation Default Language attribute and drag it to your card.Select Save Changes.You will now see the conversation's default language in the card.Any snippets used to reply to this conversation will be automatically translated to the default language, if available. See Reuse text with snippets for more information.
  • Deflection report

    Knowledge base deflection lets users self-serve common inquiries, allowing customers to find a quick solution to their questions without waiting to speak to a support agent. This reduces the volume of incoming conversations and lets your agents focus on the requests that need their attention the most.The Deflection report helps you understand how well your deflection efforts are performing and help identify any improvements that can be made to your articles to improve their effectiveness.Who can access this feature?User typesManagers can access reports.In this articleWhat is deflection?Deflection report metricsDeflection Success Rate chartArticle Performance tableDeflection Topic Coverage chartSearch Queries table To access the Deflection Report, select Reporting , and then select Deflection.Notes: The AI Response tab is only available if you use a conversational assistant with the AI Response interaction.If you do not have a conversational assistant using the AI Response interaction, this report will be titled Article Deflection. What is deflection?When it comes to customer service inquiries, deflection refers to minimizing the number of inquiries that reach a human agent by guiding customers towards resources that can help them self-serve, such as your knowledge base. Providing customers with these resources enables them to independently get the answers they need, saving time for them and your agents and enhancing their overall customer experience with your organization. At Kustomer, you can encourage customers to self-serve by setting up article deflection in your conversational assistant, emails, or forms. Organizations on Enterprise or Ultimate plans can choose to set up deflection using AI, where the conversational assistant can provide AI-generated answers for customer inquiries using information from your public knowledge base. To learn more, see Use AI to respond to customer inquiries.Deflection report metricsThe report shows 4 metrics that let you see how successful your deflection efforts have been.The following metrics are shown:Count of DeflectionsArticle deflection tab is the total count of conversations that were deflected within the selected time range and did not require resolution by an agent.AI response tab is the total count of conversations where an AI response was given, and the conversation was either marked as helpful 👍🏼, the customer abandoned the chat or the chat was marked as ended, or the conversation was not transferred to an agent. Customer Hours Saved is the total amount of time your customers saved awaiting a resolution within the selected date range. This value is generated by multiplying your organization's Count of Deflections by your organization's Median First Resolution time.Agent Hours Saved is the total amount of time saved by your team members within the selected time range. This value is generated by multiplying your organization's Count of Deflections by your organization's Average Conversation Handle time.Deflection Success Rate is the ratio of successful deflections over all conversations where deflection was attempted.Deflection Success Rate chartThis chart shows a comparison of the Deflection Success Rate metric versus the amount of Deflection Attempts, which is the count of unique conversations where deflection was attempted. The following metrics are shown:No AI Response/Article offered is when the inquiry topic entered by the customer did not match any knowledge base articles or an AI response could not be generated. Whenever a conversation’s Last Deflection Status is equal to no match, it will be counted in this metric.Offered is when an article or AI response was offered in a chat, but the customer abandoned the conversation without interacting with the chatbot further; or when deflection was offered in a form. Whenever a conversation's Last Deflection Status is equal to offered, it will be counted in this metric.Unsuccessful is when articles, or an AI response, were offered but the conversation was still transferred to an agent. Whenever a conversation’s Last Deflection Status is equal to rejected, it will be counted in this metric.Successful Article Deflection tab is when articles were offered and the inquiry was resolved without handoff to an agent. Whenever a conversation’s Last Deflection Status is equal to accepted, it will be counted in this metric.AI response tab is the total count of conversations where an AI response was given and the conversation was either marked as helpful 👍🏼 , the customer abandoned the chat or the chat was marked as ended, or the conversation was not transferred to an agent. % Deflected Article deflection tab is the ratio of the count of conversations where articles were suggested, and the conversation was marked as resolved / the total amount of conversations where conversation.lastDeflectionAt fell under the range.AI response tab is the ratio of the count of conversations where an AI response was given, and the conversation was either marked as helpful 👍🏼 , the customer abandoned the chat or the chat was marked as ended, or the conversation was not transferred to an agent / the total amount of conversations where conversation.lastDeflectionAt fell under the range. Article Performance tableThis table shows a list of all of the articles that were suggested in a deflection attempt and provides data to help you understand how each article is performing.Note: This table is only available when viewing the Article deflection tab.The following columns are available:Article Name is the title of the knowledge base article that was suggested.Last Updated is the last time that the suggested knowledge base article was updated.Offered Count is the count of unique conversations where this article was suggested based on the customer's inquiry topic. Click Count is the count of times that this article was suggested and then clicked on and viewed by the customer.Resolution Rate is the count of unique conversations where this article was selected as resolving the customer’s inquiry, over the count of unique conversations where this article was offered.Note: For article deflection in chat, any suggested articles are counted as resolving if they are viewed at least once and the customer marks their inquiry as resolved.Deflection Topic Coverage chartThis chart shows the percentage of search queries that returned at least one article, or AI response, in their results.Search Queries tableThis table shows a list of the terms customers used to describe their support requests. This information can provide insight into how your customers think about your product and help you identify any gaps you may have in your current knowledge base, such as a feature that isn't documented.The following columns are available:Search Query is the text input by the customer when prompted for their support topic.AI Response is the AI-generated response provided by a conversational assistant. This column is only available on the AI response tab.Assumed resolution rate is the count of unique conversations where an AI response was given and the customer did not mark that the response was helpful using 👍🏼, the customer abandoned the chat or the chat is marked as ended, or the conversation was not transferred to an agent, over the count of unique conversations where this response was offered. This column is only available on the AI response tab.Confirmed resolution rate is the count of unique conversations where an AI response was given and marked by the customer as helpful using 👍🏼, over the count of unique conversations where the response was offered. This column is only available on the AI response tab.Search Count is the count of unique times this exact search query was entered. This column is only available on the Article Deflection tab.Articles Offered is the count of articles returned (out of a possible 3) for this query. This column is only available on the Article Deflection tab.Resolution Rate Article Deflection tab the count of unique conversations where this query was entered and marked by the customer as resolving their inquiry, over the count of unique conversations where this query was entered.AI response tab is the count of unique conversations where an AI response was given, and the customer either marked the conversation as helpful 👍🏼 , the customer abandoned the chat or the chat was marked as ended, or the conversation was not transferred to an agent, over the count of unique conversations where this response was offered.Note: While not common, sensitive information can appear in the Search Queries table. To help minimize this risk, we recommend you follow best practices and prompt users for their name and email within the first couple of interactions.
  • Knowledge base deflection in email

     With knowledge base email deflection, you can set up an auto response message that acknowledges receipt of a customer's email, and also recommends up to 3 articles that can help answer their question without them having to wait for an agent reply. This can help encourage your customers to self serve and solve their own support request, allowing your agents to focus on more urgent matters.Who can access this feature?User typesContent administrators can access knowledge base settings. Admins can create business rules.In this articleSet up your auto response emailUpdate your theme filesSet up your auto response emailUsing a business rule template, you can quickly set up your auto response email with email deflection.Note: If your Kustomer org is multilingual, make sure your NLP workflow is setting the Lang. value on conversations. For more information, see Automatic Language Detection.To set up your auto response email:Go to Settings> Platform > Business Rules.Select the Acknowledgement auto-response template.Edit the name of the template to help you distinguish it from other auto reply templates. For this example, we will name it Auto response with deflection.In the Set Actions section, select the email address the response will come from using the From drop-down menu.Optionally, if you have a specific template you want to use, select it from the Email Template drop-down menu. Edit the message body, as needed. Select Snippets & Dynamic Text and then select the Dynamic Text tab. Add the {{kbarticle.deflection.actions}}  field to your message. Using this field will reply to the sender with 3 suggested articles that match their inquiry. Each suggested article will ask the sender if the article was helpful and have an action to close the request.In the message body, select {{kbarticle.deflection.actions}} and enter fallback text. This will be what is rendered in the auto response email if there are no articles found that match their inquiry. For example, you can enter "We're sorry, no articles were found."Select Save Changes.Make sure the business rule is enabled by turning on the toggle, if not already enabled.Agents can see the articles that were presented to the customer in the email conversation and whether or not the deflection was successful. Articles shown with a blue icon indicate that the article was read.Update your theme filesIf you are using a custom theme for your knowledge base, you need to update your theme files in order to use email deflection.Note: If you are using the default theme, you do not need to follow these steps. The files will be automatically updated with these changes.To update your custom article.jsx file:Go to Settings> Knowledge Base > Themes.Select the custom theme you are using with your knowledge base.Select your custom article.jsx file.Copy and save all of the code from your article.jsx file into a text editor.Return to the Themes page and select the Default Theme.Select the article.jsx file in the Default Theme and copy all of the code from this file.Return to your custom theme and select the article.jsx file.Select and delete all of the code in your custom article.jsx file. Then, paste all of the code you just copied from the default theme into your file.Now, copy all of your original code from the text editor (step 5).Find the line of code for return and paste all of the code you just copied from the text editor after  return (.Note: In this image, you can find return ( on line 99. This may not be the same in your custom file. Also, please note that you are replacing the code that currently comes after return( with the code you just copied from the text editor.Finally, find the </main> line and enter this line before it:{this.renderDeflectionPrompt()}Your file should resemble the following:Select Save.To update your custom form.css file:In your custom theme, select form.css.Copy and paste the following block of code in your file:/*  ============================== Deflection ============================== */ .form-deflection-prompt-container { position: fixed; display: flex; justify-content: center; align-items: center; bottom: 0; left: 0; right: 0; border: solid 1px #dbdadf; background-color: #f9fafc; padding: 15px; font-size: 16px; } .form-deflection-prompt-text {     margin-right: 40px; } .form-modal-success-icon, .form-deflection-success-icon {     display: flex; justify-content: center; color: #80ce69; font-size: 30px; } .form-deflection-item-article-cta-button { font-size: 16px; padding: 10px 30px; border-radius: 3px; background-color: #2e96cd; color: white; } .form-deflection-item-article-cta-button:hover { cursor: pointer; }Select Save.Now, if a customer sends an email and it matches any of your knowledge base articles, the auto response they receive will also include articles that can help answer their question faster.
  • Knowledge base deflection in a form

    You can turn on knowledge base deflection for your forms so that users can see up to 3 recommended articles that can help answer their questions after they submit a support form. This can help encourage them to self serve and solve their own support request, which will allow your agents to dedicate time to more urgent matters.Who can access this feature?User typesContent administrators have access to the Forms page.In this article:Turn on form deflectionView a form submission in KustomerUpdate your theme filesTurn on form deflection Go to Settings> Knowledge Base > Forms.Select an existing form or select Add Form.Turn on the Deflection toggle in the Form Settings panel.Select Save Changes.View a form submission in KustomerOnce a form is submitted, it appears in Kustomer as a new conversation and all of the information that was entered in the form is shown in the message. Form submissions can be identified by the Form channel icon.You can reply to form submissions through an email. The email address that was entered with the form submission is automatically filled in when replying to the form.Update your theme filesIf you are using form deflection and have a custom theme for your knowledge base or are using an older theme that was created before deflection became available, you need to modify your form.css and article.jsx files before you can start deflecting form submissions.Note: If you are using the default theme, these updates are made automatically and you do not need to edit your files.Go to Settings> Knowledge Base > Themes.Select your custom theme.Select your custom form.css file and copy and paste the following block of code into it:/*  ============================== Modal ============================== */ /* The Modal (background) */ .form-modal { display: flex; position: fixed; /* Stay in place */ z-index: 1; left: 0; top: 0; width: 100%; height: 100%; overflow: auto; background-color: rgb(0,0,0); /* Fallback color */ background-color: rgba(0,0,0,0.4); /* Black w/ opacity */ } .form-modal-container {     width: 100%; display: flex; align-items: center; justify-content: center; } .form-modal-box {     max-width: 600px; width: 80%; position: relative; } /* Modal Content */ .form-modal-content { background-color: #fefefe; margin: auto; padding: 20px; border: 1px solid #888; overflow-y: scroll; max-height: 700px; } @media only screen and (max-width: 768px) { .form-modal-box {     max-height: 100%; max-width: 100%; width: 100%; } .form-modal-content {     height: 100%; max-height: 100%; width: 100%; -webkit-overflow-scrolling: touch; } } /* The Close Button */ .form-modal-close {     display: flex; justify-content: flex-end; } .form-modal-close:hover, .form-modal-close:focus { color: #000; cursor: pointer; } .form-modal-success-container {     display: flex; align-items: center; justify-content: center; flex-direction: column; } .form-modal-success-container-no-results {     margin: 50px auto; } .form-modal-success { text-align: center; font-size: 16px; } .form-modal-divider {     margin: 20px auto; background-color: #dbdadf; height: 1px; } .form-deflection-header {     font-size: 18px; font-weight: bold; } .form-deflection-description { font-size: 16px; margin: 15px auto; } .form-deflection-item {     border-radius: 3px; border: solid 1px #dbdadf; margin-bottom: 15px; } .form-deflection-item-article {     padding: 20px; } .form-deflection-item-article-name { font-size: 16px;     font-weight: bold; display: flex; align-items: baseline; } .form-deflection-item-article-body {     margin: 10px 0; font-size: 14px; } .form-deflection-item-article-actions {     display: flex; justify-content: space-between; align-items: center; } .form-deflection-item-article-read-more, .form-deflection-item-article-link { display: flex; align-items: center; font-size: 14px; color: #2e96cd; cursor: pointer; } .form-deflection-item-article-link {      margin-left: 5px; line-height: 16px; } .form-deflection-item-article-read-more i { line-height: 14px; } .form-deflection-item-article-read-text {     margin-left: 5px; } .form-deflection-item-article-cta { border-top: 1px solid #dbdadf; background-color: #f9fafc; padding: 20px; display: flex; justify-content: space-between; align-items: center; } .form-deflection-item-article-cta-text {     font-size: 16px; }Select your article.jsx file and copy and paste the following code below the this.handleCloseRequest = this.handleCloseRequest.bind(this); line.this.handleCloseFormHookRequest = this.handleCloseFormHookRequest.bind(this); this.handleCloseProductizedFormRequest = this.handleCloseProductizedFormRequest.bind(this);Make sure you add it between the { }.Copy and paste the following code into your article.jsx file.Note: If you are already using the handleCloserequest() function, replace it with this code.handleCloseProductizedFormRequest(articleVersionId, closeOnLoad) {          const { data, org } = this.props; const formId = _.get(data, 'deflection.formId'); const externalId = _.get(data, 'deflection.externalId'); const queryTerm = _.get(data, 'deflection.queryTerm');     const emailValue = _.get(data, 'deflection.emailValue');     const emailLabel = _.get(data, 'deflection.emailLabel'); const url = `https://${org.name}.api.kustomerapp.com/p/v1/kb/forms/${formId}/submissions`; const formData = { externalId, }; const deflectionData = { status: 'accepted', messageSubject: queryTerm, articleVersionId, }; const customerEmailData = {     value: emailValue, label: emailLabel, attributeType: 'customer', attributeProperty: 'emails' }; formData.deflectionTracking = JSON.stringify(deflectionData); formData.customerEmail = JSON.stringify(customerEmailData); return fetch(url, {     method: 'PUT', body: JSON.stringify(formData), headers: { "Content-type": "application/json" } }) .then(() => {     if (closeOnLoad) { this.setState({ closedRequest: true }); } }) .catch((err) => { console.error(err); }); } handleCloseFormHookRequest(articleVersionId, closeOnLoad) {     const { data, org } = this.props; const externalId = _.get(data, 'deflection.externalId'); const conversationId = _.get(data, 'deflection.conversationId'); const signature = _.get(data, 'deflection.signature'); const source = _.get(data, 'deflection.source'); const orgId = _.get(org, 'settings.org'); let url = ''; const patchData = {}; if (externalId) { url = `https://${org.name}.api.kustomerapp.com/v1/hooks/form/deflection/${orgId}/${externalId}/${articleVersionId}`;     patchData.status = 'accepted'; } else if (conversationId) { if (closeOnLoad) { url = `https://${org.name}.api.kustomerapp.com/p/v1/kb/conversations/${conversationId}/deflection`;     patchData.status = 'accepted'; } else { url = `https://${org.name}.api.kustomerapp.com/p/v1/kb/conversations/${conversationId}/visit`; } } if (conversationId && source === 'email') { patchData.org = orgId; patchData.articleVersionId = articleVersionId; patchData.signature = signature; } return fetch(url, {     method: 'PATCH', body: JSON.stringify(patchData), headers: { "Content-type": "application/json" }     }) .then(() => {     if (closeOnLoad) { this.setState({ closedRequest: true }); } }) .catch((err) => { console.error(err); }); } handleCloseRequest(articleVersionId, closeOnLoad) { const { data } = this.props; const formHookEnabled = _.get(data, 'deflection.formHookEnabled'); if (formHookEnabled && formHookEnabled === 'false') {     return this.handleCloseProductizedFormRequest(articleVersionId, closeOnLoad); } return this.handleCloseFormHookRequest(articleVersionId, closeOnLoad);       }Select Save.Now, when a customer uses your form to submit a support request, they will get a list of suggested articles once they submit the form. Notes: Articles must be assigned to a category in order for them to be presented as a suggested article.Articles are suggested based on what your customer input in the Subject line of the form.The customer can select any of the suggestions to view the article in its entirety. If it resolves their request, the conversation that was created by the form in Kustomer will automatically close once they select Yes, close my request.
  • Agent Suggestions

    Agent Suggestions help improve your agent productivity by using natural language processing (NLP) techniques to automatically analyze messages and suggest the most likely shortcuts that can be applied to a conversation. Shortcuts can contain canned responses and perform actions on a conversation.August 2023 update: This feature is no longer available to new customers.Who can access this feature?Admins can turn the feature on or off.Agents who have Read access to Shortcuts via permission sets can apply suggestions to conversations.In this articleTurn on Agent SuggestionsApply suggested shortcuts to conversationsSearch and report on agent suggestionsTurn on Agent SuggestionsYou can turn on the Agent Suggestions feature for your organization by going to Settings > Kustomer IQ > Agent Suggestions.By default, all users with access to shortcuts have access to agent suggestions when you turn the feature on. You can limit access to only specific users and/or teams by selecting them in the available fields. Apply suggested shortcuts to conversationsOnce the feature is turned on, Kustomer is able to predict shortcuts that your agents can use when replying to email conversations.Notes:Agent Suggestions are only shown on first inbound email messages.Once a conversation is created within Kustomer, it will take approximately 5 minutes before suggestions are made.The last message must either have come from the end customer or was automatically sent by the system. For example, an auto response message that was sent by a business rule.Once an email message is received, the system presents an agent with up to 3 shortcut suggestions.You can hover over a suggestion to see if the shortcut contains a canned response, actions, or both. In this example, the shortcut consists of a canned response, an action that changes the subject of the conversation, and one that applies relevant tags to the conversation.If you select a suggestion a window opens that shows you what message or actions will be applied to the conversation when using that shortcut. Select Apply message and all actions to apply the suggestion.Tip: You can bypass this view by selecting the Don't show again checkbox. Doing so will automatically apply the suggestion when you select it in a conversation.In this example, the shortcut changed the conversation title, added tags, and inserted a message into the reply window.Note: Once you apply a suggestion or start typing, the list of suggestions will hide. You can view them again by selecting the icon that is located at the bottom right of the reply window.Search and report on agent suggestionsYou may want to create search queries or build custom reports to view how often suggestions are being offered or who is applying them the most. You can use one of the following values to create your filters and get a better idea of how agent suggestions are being used in your organization:Suggested Shortcuts shows the conversations that were offered suggestions.Suggested Shortcuts Applied At shows conversations that were applied in the time frame you specify using the available operators. To use this value when building a custom report, select Conversation Suggested Shortcuts Applied from the data attribute drop-down menu.Suggested Shortcuts Applied By shows the conversations where suggestions were applied by a specific user.The Suggested Shortcuts and Suggested Shortcuts Applied By values are also available for you to use when segmenting your custom report so that you can get a better idea of how your data compare against each other. For example, you can segment conversations where suggestions are available and see in which conversations the shortcut was or wasn't applied. Note: These options are only available for customers with the KIQ for Agents plan when searching or reporting on the Conversation object. For more information on available operators, see Understanding filter operators.
  • Best practices for creating custom AI prompts

    With AI Profiles, you can customize how AI response and Agent Assist respond to customer inquiries and ensure they align with your brand's individual tone and style. You can also provide guidelines on how you want responses to be formatted. This article provides tips to consider when creating your custom prompts.See Tailor AI responses to customer inquiries with AI Profiles to learn more about this feature.Who can access this feature?User typesAdmins can access the AI Profiles page.In this articleDescribe your company and productDefine clear objectivesDescribe the tone you want the AI to useSpecify content structurePreview and iterateDescribe your company and productProviding this detail can help the AI personalize responses more effectively and make interactions more engaging. When describing your product and company, you can include:The company name and type of product(s) you offerBenefits of the product(s)A brief description of your company and its missionExample: You are a customer service chatbot for Acme, a retail company specializing in camping equipment. Our equipment can be used by any camper and is tailored for individuals new to this activity due to its simplicity. We want to make camping fun and less intimidating for those who want to start learning about it. Be knowledgeable about all Acme products and provide information about availability, prices, and materials.Define clear objectivesBe as thorough as possible. The more details you give, the better as it helps the AI model better understand your expectations. Be specific about how you want the AI to respond. Use grammatically correct sentences when writing your prompts. Examples: If you cannot find an answer to the question, say 'Would you like to talk to an agent?' instead of 'Contact us'.If a customer asks a question about a delayed order, apologize for the delay and provide a clear and concise explanation.Always offer a specific solution or next steps to resolve the customer's issue.Describe the tone you want the AI to useClearly define the desired tone for AI responses using descriptors like professional, casual, or friendly.Examples:Always respond friendly and approachable and make customers feel welcomed and excited about our products.When responding to a customer’s concern, be professional and formal to demonstrate that we take their issue seriously.When sharing information about a current promotion, always respond enthusiastically and try to generate excitement over the upcoming deal.Specify content structureInclude the preferred structure for responses in your prompt customization. Examples:Make sure to always greet the customer based on the time of day.When writing conversation summaries, limit the summary to no more than 5 bullet points.Preview and iterateIterating on your custom prompts is the key to your success. While the prompt length can vary, we found instructions ranging from 100-200 words to be the most effective in our testing. While creating a new prompt, we encourage you to use the Preview feature regularly to view how the outputs change and make changes as necessary.Once your profile is live, periodically review your conversations to see if the AI-generated responses suit your needs, and iterate on the prompts, if necessary.
  • Tailor AI responses to customer inquiries with AI Profiles

    Use AI Profiles to create custom prompts that ensure responses generated by Customer Assist Premium or Agent Assist align with your brand's unique identity. This feature allows you to define specific tones, styles, and guidelines for how you want these features to respond to customer inquiries. By tailoring these responses, you can maintain a consistent voice across your communications, build trust by providing more accurate answers, and elevate the overall customer experience. Who can access this feature?User typesAdmins can access the AI Profiles page.In this articlePrerequisitesCreate a new profileDefine your custom promptsActivate the profilePrerequisitesYou must have access to Customer Assist Premium and/or Agent Assist to create AI Profiles. For more information on these features, see:Use AI to respond to customer inquiriesBoost agent performance with KIQ Agent AssistCreate a new profileGo to Settings > KIQ > AI Profiles and select Add profile.Decide if you want to create a new profile for Customer Assist or Agent Assist.Enter a name and select the language in which you are writing the prompts. The following languages are supported:LanguageAgent AssistCustomer AssistEnglishXXFrenchXGermanXItalianXJapaneseXBrazilian PortugueseXXSimplified ChineseXSpanishXXOptionally, select a brand if you are creating a profile for Agent Assist. You do not need to select a brand if you are creating a profile for Customer Assist.Note: For Agent Assist, you should always assign a brand to your profile unless you are creating a profile that will be used as the default profile across your organization. You can set a profile as default by going to Kustomer IQ > Agent Assist and selecting the prompt from the Default AI Profile menu.If you are creating a profile for Customer Assist, select the assistant(s) for which you are creating the profile in the drop-down menu. Only published assistants using the AI response interaction will appear in this list.Define your custom promptsAdd your custom prompt instructions in the provided field. If creating a profile for Agent Assist, you can create a different prompt for each feature by selecting each tab.For more information, see Best practices for creating custom AI prompts.You can preview how the AI will form responses using your prompts in the Preview section for each feature.Note: When testing a response for the Generate response or AI response features, make sure your knowledge base can answer your example question.Once you have your final prompts, Save your changes.Activate the profileOnce you save your changes, use the toggle to activate the profile on the main AI Profiles page. If the profile is also used across your organization as the default for Agent Assist, you will see a Default tag next to it.
Powered by Kustomer