Discuss ideas, submit feedback, and track what's next
Help us understand your request
An intake-ready template
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)Hello, I started to work with AI prompting and try to use it in Action Builders, AI Agents, App Builders..As you know, AI needs a lot of test before beeing set in production, and some tests means review a prompt (add a new instruction) trying to optimise Agent Behaviour. What problem do you see this solving? (1-2 sentences)The problem I meet is that on some complex AI Agents with a lot of prompts, at some time the systems start to hallucinate or the new addings I put are not fully applied.To solve this issue, maybe an option may be to have a feature that can “Reprompt” in one unique prompt all the work done, or something like : what is the simplest prompt to arrive to this result. This simplest can be then used to go deeper in the project.. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)some days ago with all AI I tested (Chat Gpt, Claude, Zendesk…, Copilot..). At a point, it seems easier not to continue to argue with the AI but to restart from scratch with the experience learned in first issues. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)I tried sometimes to copy paste all my prompts in our AI internal solution to get a clean prompt, but not working smoothly. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)As I told maybe a capacity for AI to look in the app or agent built by prompting and from this reconstrut the simplest prompt to arrive to the same result.
When customers in the Help Center use the AI customer support chat box, the box takes up the entire screen on mobile devices. When customers click on a link the AI recommends (ex: specific help center article or submit a request page), it will take them to that page but then the chat box cover the entire screen which makes it feel like there was an error and the link didn’t actually work. Clicking the minimize chat button enlightens the customer that they were navigated to the correct page but it doesn’t seem to be obvious enough for some customer. I was wondering if having the chat box only take up ~80% of the screen (enough to be able to see the top of the page one is currently on) would be enough of an indicator that they need to minimize the chat box. Thank you for your time.
Hi,I'm writing to regarding an urgent issue related to a GDPR non-compliance that exposes us to penalties and user issues.The main problem is that AI Agents tickets cannot be edited either by admins, or with triggers/automations, and even with API calls, and if a user requests an account cancellation it remains blocked with this error:"The user is registered as a requester for 1 ticket that is not yet closed" So the automatic flow of account deletion is blocked, and we should always proceed with manual deletions, waiting for the automation to go on fire, but we have millions of users, and our team that handles these topics can't manually delete each user individually (with possible human errors).We already opened a ticket with Zendesk Support, but it did not provide us with definitive solutions, just to edit the automation, and they said to open a post here to give a feedback to Zendesk Product team.We edited the default Zendesk automation (the only one that works for ticket from chatbot AI) reducing the hours to change the ticket status from solved to closed, but this doesn't solve the problem.We need to address this issue upstream:regardless of the ticket status, the AI ticket absolutely must not be blocking for user account deletion.We have been using the chatbot since 18/06/2026 and have already had 2 cases in a few tickets.So, unfortunately, we had to turn off the AI Agents chatbot because our Legal team wants to prevent further cases of GDPR non-compliance.SOLUTION: The solution is to make tickets from the AI chatbot editable like other Support tickets or at least allow users account deletion at any time, even if they have an ongoing ticket with the chatbot.
Hello!We would like to request a configurable reply delay for AI Agent responses on the WhatsApp channel. We use the AI Agent to answer customer questions via WhatsApp based on the content in our Zendesk Help Center. A configurable reply delay would allow the AI Agent to wait until customers have finished sending their messages before generating a response. Use caseMany customers naturally send their question in multiple WhatsApp messages instead of typing everything in a single message. For example:"Hi" "I have a question about my order" "Order number is 12345" "The issue is..."Currently, the AI Agent responds immediately after the first message, which often results in incomplete context and unnecessary back-and-forth conversations.Requested functionalityIt would be extremely valuable to have a configurable reply delay (minutes) before the AI Agent generates its response. During this delay, additional incoming WhatsApp messages from the same customer would be combined into a single request, allowing the AI to respond with the full context.BenefitsBetter understanding of the customer's complete question. Higher quality AI responses. Fewer unnecessary follow-up questions. More natural conversations that align with how people typically use WhatsApp. Improved customer experience and reduced AI token usage by avoiding multiple partial responses.I understand there is currently no native support for this functionality, but I believe this would be a valuable enhancement for many customers using AI Agents on WhatsApp and other asynchronous messaging channels.I hope the Product team will consider adding this feature in a future release.Best regards, Team SportBit
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)We are requesting that Zendesk Agent Copilot Auto Assist be updated with multimodal vision capabilities so it can read and analyze ticket attachments like photos, screenshots, and videos. This limitation directly impacts our front-line support agents who rely on customer-uploaded media to troubleshoot, as well as our admins who design Auto Assist Procedures. Adding vision support will ensure Auto Assist maintains context when critical customer data is sent as an image rather than text. What problem do you see this solving? (1-2 sentences)This feature solves the context gap that occurs when Auto Assist falls "one step behind" because it cannot process customer-submitted images, such as product serial number labels or error screenshots. By allowing Auto Assist to ingest visual attachments, the AI can seamlessly complete procedural workflows that depend on visual verification without breaking the suggested reply flow. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)This issue affected us today and continues to impact approximately half (50%) to two-thirds (66%) of our daily incoming support tickets where serial number photos are requested. When a customer sends a photo of their product label, Auto Assist misses the visual details entirely, proceeds with the wrong text-only steps, and throws off the procedural flow. This constant disruption forces our agents to frequently dismiss or completely disable Auto Assist, which dramatically slows down handling times and reduces our overall return on investment in the Agent Copilot suite. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)Yes, our current workaround requires agents to manually download the customer's attachment, open the image to visually extract the product details, and manually type a reply based on that information back into the ticket or custom fields. Because Auto Assist is blind to the attachment and falls out of sync, agents have to dismiss the AI and take over the interaction entirely. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)Our ideal solution is a multimodal Auto Assist where the underlying model can detect and process visual media attachments via instructions built directly into procedures (e.g., "Analyze the attached photo to extract the serial number and model"). Auto Assist would then automatically extract the information, populate the corresponding custom fields, and suggest the correct next troubleshooting step without breaking stride.
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)We have received some feedback from agents that the AI summary view section is rather limited and is hard to scroll and read for summaries of long ticket threads What problem do you see this solving? (1-2 sentences)Agents getting up to speed on assigned tickets faster, reduced clicks are scrolling to navigate the UI. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)No workaround, agents are scrolling the summary as needed What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)Allow users to expand the AI ticket summary as desired
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)Two enhancements for the AI Agent Advanced: first, the ability to configure up to three sequential inactivity triggers with distinct re-engagement phrasing; second, the ability to trigger an automatic session closure message after 2 hours of inactivity, similar to how Inactivity triggers currently work. This affects our admins, who cannot build optimal conversational flows, and our customers, who miss out on a proactive and clear conversational experience. What problem do you see this solving? (1-2 sentences)This will solve the high drop-off rates (most of all whatsapp) by allowing more persistent and varied re-engagement attempts before a chat is abandoned. Additionally, it ensures a transparent customer experience by clearly communicating to the user when their session has been automatically closed due to time-outs. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)This is an ongoing daily issue, especially on the WhatsApp channel, where users frequently close the app, switch to other chats, or get distracted and forget about the open conversation. Currently, sending just one inactivity message is not enough to win back their attention, and because sessions close silently after 2 hours, users often try to resume the chat later expecting their previous context to be active, only to find a dead session. This lack of transparency and lost context causes significant customer friction, leading to confusion and lower bsat scores. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)No, there is no viable workaround within the AI Agent Advanced to cycle multiple different inactivity messages or to force a final text notification exactly at the 2-hour automatic session closure. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)The ideal solution is to allow admins to stack up to 3 inactivity steps in ai agent advanced (e.g., after 15, 30, and 45 minutes) each with its own customizable text behavior. Additionally, a native "On Session Close" trigger should be added to the AI Agent settings to automatically send a final goodbye/closure message after the 2-hour timeout.
SummaryZendesk's Intelligent Triage assigns an intent to incoming tickets based on subject line and first public comment. When the classification is wrong, agents can manually correct it. Those corrections have no effect on future classifications. The model does not learn from them.For accounts where intent accuracy matters -- routing logic, prioritisation, Explore reporting -- this creates a ceiling on how good the system can get. The only path to improvement is ongoing manual curation of intent definitions, which doesn't scale.The gapIntelligent Triage is positioned as a system that classifies tickets automatically, reducing manual handling. In practice, the accuracy floor is set at initial configuration and stays there. Agent corrections are a rich, continuous source of ground-truth signal and currently none of it feeds back into the model.We've used the quality recommendations feature, worked through intent definitions, and reduced overlap. After all of that, we're still seeing a meaningful volume of misclassified tickets. The tooling for improving intent definitions is reasonable. The problem is that corrections made at the ticket level -- the most direct signal available -- are simply discarded.This was raised in the Intelligent Triage article comments in 2022. The response at the time was that the system didn't learn from agent feedback "at the moment." Three years on, confirmed with Zendesk support, the answer is the same.What we're asking forAgent corrections to the intent field should be used as a training signal to improve future classifications. This doesn't need to be real-time or fully automated. A supervised process where accumulated corrections periodically inform the model would be sufficient. The data already exists every time an agent overrides a classification.Who this affectsAny account using Intelligent Triage at scale where intent accuracy is load-bearing -- routing automations, SLA triggers, or reporting built on intent as a dimension. The current model works reasonably well at launch and degrades in usefulness relative to the effort required to maintain it.
Zendesk just renamed Intents (Intentions) as Topics (Sujets).The weird part of it is that in French it’s now the same word for the Subject and the Topic, which is confusing.This can create confusion at the moment of checking ticket subject or ticket topic. We can’t use a workaround about this issue because it’s a system configuration. Ideal solution would be to replace the translation of Topic Intention : straightforward, and same as before. In the AI and NLP (Natural Language Processing) world, this is the gold standard. ⭐Thématique : would be my favorite, this is exactly what you were trying to translate in the first place, for me the best translation to Topic Motif : This is the operational, CRM-focused term used by support teams and managers, could be translated by ‘Contact reason’. I would save this for ticket fields.Happy to have a chat, I’ve already raised some time ago the bad translation of ‘Email’ in french (Messages) and it has been modified since.
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)I would like to request the ability for Pro Tiers with Email AI to automatically resolve Zendesk tickets when a customer confirms their issue has been resolved, either through a written response or by clicking a confirmation button within the email. This would primarily benefit agents and administrators by reducing manual ticket management, while also providing customers with a simple way to confirm resolution. Notably, a similar capability was available in AI Essentials, so, surprisingly, it is not currently available. What problem do you see this solving? (1-2 sentences)Many tickets remain open even after customers confirm their issue has been resolved, requiring agents to manually review and close them. Automating this process would reduce unnecessary administrative work and improve ticket lifecycle management. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)While this has not yet occurred in our current workflow, we identified this gap while setting up Email AI and anticipate it becoming a recurring issue. As we scale usage, customers will inevitably confirm resolution via email, but tickets will remain open until an agent manually updates their status. This will create additional work for agents who must monitor resolved conversations solely to close tickets. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)Currently, agents must manually identify customer confirmations and update ticket statuses themselves. While effective, this adds unnecessary manual effort and does not scale efficiently. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)Pro tiers with Email AI should automatically detect clear customer confirmations that an issue has been resolved and update the ticket status to Resolved. Alternatively, or in addition, customers could be provided with a "Confirm Resolution" button in email communications that automatically resolves the ticket when selected.
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)We would like Zendesk to add the ability to drill down into conversations where a specific Knowledge Source article was used by the AI Agent. Currently, Knowledge Source analytics only provide aggregated metrics such as Usage Rate, Automated Resolutions, Verified Resolutions, and BSAT, without any way to inspect the underlying conversations. This limitation primarily affects AI administrators, knowledge base managers, customer support leaders, and content owners responsible for maintaining and optimizing AI performance.What problem do you see this solving? (1-2 sentences)This functionality would enable teams to understand how individual knowledge articles perform in real customer interactions and identify content gaps, inaccuracies, or opportunities for improvement. It would also make Knowledge Sources as actionable and manageable as AI Use Cases, which already support conversation-level analysis.When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)We encounter this limitation regularly during our AI Agent performance reviews and knowledge base optimization initiatives. For example, when an article shows a low BSAT score or a high escalation rate, we have no way to understand what happened in the conversations where that article was used. Without access to those interactions, it is difficult to determine whether the issue stems from the article content, customer intent mismatches, or AI behavior. This occurs on an ongoing basis and significantly slows down our ability to improve automation performance, customer satisfaction, and knowledge base quality.Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)There is currently no effective workaround. We can only review high-level metrics and attempt to manually identify related conversations through separate reporting or ticket searches, which is time-consuming and often inconclusive.What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)Each Knowledge Source article in the analytics view should be clickable, similar to AI Use Cases, allowing administrators to view and filter all conversations where that article was referenced by the AI Agent. Ideally, knowledge sources should be added as a separate filter in Conversations Logs.
Hi,I set up a custom agent to extract the serial number of an image.This works when a logged-in user submits the request and attachment via the form, but not when the attachment is sent via email.This is the error I get:Note: The attachment "WhatsApp Image 2026-05-08 at 15.37.14 (1).jpeg" could not be downloaded because it is hosted on support.company.com (which is the hostmapped help center URL), which does not match the Zendesk tenant domain. Please manually review the attachment to extract the serial number and model number.”I guess this has to do with authentication, but is there a solution for this?
It would be incredibly helpful to have an AI review all tickets related to an organization, specifically to identify: What was the issue? Was it resolved? Was support good? Was support bad? What was the customer sentiment? Currently we have to ask a team member to manually audit all the tickets, but it would be great to have this automated.
Please give a quick overview of your product feature request or feedback, and note who in your organisation is affected by this issue [e.g. agents, admins, customers, etc.] Provide a supported way to set ticket form, group, and custom fields on AI-agent-only messaging tickets (native_messaging) that resolve without escalating to a human, scoped dynamically per bot, procedure, or tag. This affects admins (who can't configure workflows for these tickets), agent teams and team leads (who rely on correct form and group routing), and operations/reporting staff (who need accurate Explore data). In our large, multi-brand instance, it touches every team we put on AI messaging. What problem do you see this solving? It would let bot-resolved conversations carry the correct form, owning group, and source field so they can be routed and reported on accurately, instead of landing as read-only tickets we can't classify. It would also let AI messaging scale to multiple teams that each require a different form. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur, and how does this impact your business? This affects us daily, on every chat our bots fully resolve, which is a large and growing share of our messaging volume. Most recently, when a team asked us to log their AI-resolved chats to a dedicated form with the right group and a "Ticket source" field, we found no supported path: triggers and automations don't run on these tickets, and the AI agent's Update ticket info action can't set the form. The only lever, reordering the account-level forms list, would force one form onto all AI tickets and break the triggers and automations built on our current default, so it's unusable for us. The impact is that an extensive and increasing volume of AI-resolved tickets can't be classified into the correct form, and because the tickets are read-only, they can't be corrected afterwards either, leaving a significant blind spot in our volume and CSAT reporting. Are you currently using a workaround to solve this problem? (If yes, please explain) For group and custom fields, yes: the AI agent's Update ticket info CRM action can set them during the conversation. For the ticket form, there is no workaround at all, because these tickets are fully read-only: not even an admin can change the form (or anything else) after the fact, so the form stays permanently incorrect on every AI-resolved ticket. What would be your ideal solution to this problem? How would it work or function? Add "ticket form" as a settable field in the AI agent's Update ticket info CRM action, alongside the existing group and custom-field support, scoped per bot, procedure, or tag, and applied at ticket creation without requiring escalation. Reusing the mechanism that already sets group and fields keeps it consistent and avoids depending on triggers, which don't run on these read-only tickets.
The ask: Auto Assist: When we choose to ‘edit’ an Auto Assist suggestion to tweak it, then while we are in ‘Edit’ mode' if a customer updates the ticket simultaneously, the content we edited will be lost. We can't save the edited content. So, if an agent adds major (relatively time consuming) things to the suggestion, then as soon as the customer update is added to the ticket, suddenly ,the edited content vanishes, and it reverts to the original content that was suggested initially. The above is a big pain if the customers update was insignificant (eg. correcting a typo) and if the agent has spent a lot of time editing the suggested answer. Further background information: We see this as a bug which needs to be fixed. However, In our ticket 13679816, Zendesk support say that this is working as designed. They say that Auto Assist is designed to generate a new suggestion in the composer each time an end user sends a new message. This resets the suggestion area with a newly generated reply based on the most recent customer context, as outlined in the product documentation: Auto assist provides a new suggestion after every customer response.This means that if an agent is actively editing a previous suggestion and the end user replies again before the edit is saved or approved, the in-progress edits will be lost as the suggestion is replaced by the new one. We apologize for any inconvenience about this.Regarding feedback and plans for changing this, Zendesk acknowledge that this can be really frustrating to users especially when agents are in the middle of editing and their edits are overwritten by a new suggestion. However, as of the latest available information, there are no details yet on their roadmap regarding a planned change to this workflow. Any related improvements, such as support for retaining or recovering in-progress edits, or alerting agents before regenerating, have not been formally announced yet. They advised me to raise this Community Post to request that this functionality be improved.
Hi Zendesk Team, Our agents frequently need to attach files or paste images when responding to customer inquiries. While they appreciate the efficiency of AI Auto Assist, they currently have to disable it and manually compose their response whenever an attachment is required—this happens in approximately 25% of our customer replies. To fully leverage Zendesk AI and streamline our support workflow, it’s essential that AI-assisted replies support file uploads and image pasting. This enhancement would significantly improve our team's efficiency and ensure a seamless experience for both agents and customers. We would love to know if this feature is on your roadmap. Thank you for your time and consideration!
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.]I am requesting the ability to "pause" a procedure in the Advanced AI Chat Bot to link to another Use Case (Procedure or Dialogue) at any point, and then return to the original procedure once complete. For example, the bot could pause the current procedure, run another use case to display a dynamic carousel and collect a user parameter, and then pass that context back to "unpause" the first procedure and continue the flow. This primarily affects admins and bot builders in our organization.What problem do you see this solving?Currently, bot builders have to design procedures that "jump" to different actions based on parameters collected in future steps, and then awkwardly link back to the original procedure. This feature would eliminate these convoluted jump loops, solving major design and maintenance headaches when building or updating bot capabilities.When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business?I am affected by this every time I need to build even a semi-complex procedure. Because we cannot simply pause and return, I end up spending hours caught in a tedious cycle of building, testing, rebuilding, and testing again. This is a massive drain on productivity and significantly slows down our ability to deploy new bot functionalities.Are you currently using a workaround to solve this problem? (If yes, please explain)Yes, the current workaround is to update the procedure with a "check step." I have it check if a specific parameter exists; if it does, it skips the first part of the procedure that calls the use case. If it doesn't, it continues to the step that collects the parameter. This is incredibly inefficient—if you need to collect 2-3 parameters and build different flows for each, it quickly devolves into a highly complex, unmanageable spiderweb of routing.What would be your ideal solution to this problem? How would it work or function?The ideal solution is to allow procedures to naturally "Pause" when linking to another use case and automatically "Unpause" (return to the exact point it left off) when that use case is complete. The procedure could then continue normally using simple branching steps, entirely removing the need for a procedure to be called over and over again from the beginning.
Today, when working with Zendesk AI Agents and JSONata, it is possible to retrieve and display data from APIs (including large arrays), but there is no way to dynamically reference an array position based on user input during the conversation.A very common use case is displaying a numbered list to the user, for example: 1 - Option A 2 - Option B 3 - Option C ...The user then replies with a number (e.g., “84”), and the expected behavior would be to map this input back to the corresponding item in the array and retrieve its attributes (such as an ID).However, today it is not possible to perform something like: var_array[user_input - 1].idThis creates a significant limitation when building scalable conversational flows that rely on dynamic data from APIs.Why this mattersMany APIs return large and dynamic datasets (100+ items, growing over time) Index-based selection is a simple and intuitive UX pattern for users There is currently no native way to map user input back to stored array data JSONata cannot be re-evaluated dynamically during the conversation using user inputAs a result, developers are forced to:Use external middleware (e.g., APIs, automation tools) to resolve the selection Or redesign the experience (e.g., forcing buttons or embedding IDs), which is not always feasibleLimitations of current workaroundsCreating individual variables for each item does not scale and breaks as data grows External processing adds latency, complexity, and dependency on additional infrastructure UX alternatives are sometimes restrictive or not suitable for all channelsSuggested improvementEnable dynamic evaluation of variables using user input within the conversation context.For example:Allow expressions like var_array[user_input - 1] Support runtime evaluation of JSONata (or similar) using variables collected during the dialogue Provide safe handling for invalid indexes (out-of-range, non-numeric input, etc.)Expected benefitsFully dynamic and scalable conversational flows Reduced dependency on external systems Simpler implementations for common patterns (lists, selections, lookups) Better developer experience when working with APIs More natural and flexible user interactions
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)We frequently have to add CCs to our public replies for customers due to expectations from them to include supervisors/etc. Right not, we have to turn off the auto assist, add the cc, go back through and turn it back on before we are able to send with a cc. What problem do you see this solving? (1-2 sentences)Faster workflow and less clicks for agents that are needing to handle interactions. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)Daily Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)Manually turning off the AI and turning it back on. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)Somewhere on the auto assist that allows to add the cc without having to go turn everything off and back on.
Our agents don't find the behavior of the Approve button intuitive. When we introduced Auto Assist, many of them accidentally sent a reply to the customer before they were ready at least once, because they are so used to only the Submit button sending the reply. We would prefer if there were only the option to “Accept” the suggested text and actions, then the actions are applied (not performed - similar to macros), and the text is copied to the regular editor. This would also avoid awkward situations where Auto Assist is correctly suggesting to set the ticket to solved, but the ticket still has an empty drop-down for a field that is required for solving. Agents first have to fill the required field, submit the ticket, and then approve Auto Assist. If they forget, Auto Assist won't perform the solving action.Edit to add: When agents edit an auto-assist reply and then accidentally click “submit as pending” before approving the text, their reply will be gone and can't be restored. This is especially frustrating when it was a longer text.cc Andreas Ackermann Johannes Lober
Since for humans an image can say more than a thousand words our articles contain images. An AI LLM can however not read and/or understand images. Why is your article editor not informing your customers of this huge problem. Please implement an automatic request by the system for the user to ad an html (text) placeholder and the best implementation would actually be 1 placeholder for the customer and another to explain the image to the AI implementation (which will clearly be quite difficult).
We would like to provide feedback regarding the current behaviour of the Co-pilot response box, specifically related to the visibility and editability of the To and CC fields. At present, the To and CC fields are not visible or editable within the response interface. This limitation prevents agents from reviewing or amending recipients in accordance with our internal operational procedures. As a result, we contacted Zendesk Support and were provided with a workaround. However, the workaround requires agents to complete the following steps: Turn off Auto-Assist Manually remove the CCs from the ticket Click Submit without making any other changes Re-enable Auto-Assist Accept the Co-pilot–generated response While this workaround allows agents to amend the CC fields, it introduces multiple additional steps into the ticket-handling process. This significantly reduces efficiency and undermines the intended purpose of Co-pilot as a productivity-enhancing tool. More importantly, this limitation introduces operational risk. Without clear visibility and control over the To and CC fields during the response process, there is an increased likelihood of responses being sent to unintended recipients without proper agent verification. We hope the team will consider this feedback as part of ongoing improvements to Co-pilot. Even if this impacts a limited number of customers, recipient visibility and control are critical for maintaining accuracy, compliance, and trust in customer communications. Thank you for considering this enhancement.
In my enviroment I have to create a segment and search rule for each organization (client) and I have over 100. With each AI Agent I have to duplicate these. It would be nice if I could just copy the AI agent or the segment/search rules from one to the other. If I am missing it, just let me know, thanks.
Please give a quick overview of your product feature request or feedback and note who in your organization is affected by this issue [ex. agents, admins, customers, etc.] (2-3 sentences)Our agents have different processes based upon the channel that they are using to communicate with a customer. Example: Live messaging you can ask questions back and forth rapid fire, whereas in email we try to get all of the questions answers all at once for troubleshooting. What problem do you see this solving? (1-2 sentences)This would help us to focus procedures based upon the type of interaction that is not available now. When was the last time you were affected by this lack of functionality, or specific tool? What happened? How often does this problem occur and how does this impact your business? (3-4 sentences)This affects our support agents daily as the procedures cant be focused by channel accurately and may lead to additional contact points with the customer when they dont need to be. Its causing more back and forth interactions asking for clarifying information that it should. Are you currently using a workaround to solve this problem? (If yes, please explain) (1-2 sentences)We are only able to utilize the auto assist feature with our messaging channel at the moment because its causing so much additional work with our email and web form channels. What would be your ideal solution to this problem? How would it work or function? (1-2 sentences)If we could somehow set it to only fire with specific channels flows, it would allow us to be able to function with our support center better. With intent, its the same across both channels and an additional filter would solve our issue.
Nosso time de Melhoria Contínua, em conjunto com a área de Qualidade, identificou um comportamento no fluxo do BOT AI via WhatsApp que pode impactar negativamente a experiência dos membros.Observamos que, após o encerramento de uma conversa, alguns membros retornavam contato pelo mesmo canal. Entretanto, essas novas interações não eram direcionadas automaticamente para um atendente humano caso o membro não selecionasse corretamente uma das opções do fluxo de triagem.Quando o membro enviava mensagens livres, sem interagir diretamente com os botões apresentados pelo BOT, o agente de IA encerrava automaticamente a conversa, fazendo com que alguns casos permanecessem sem atendimento adequado.Esse cenário pode gerar percepção de falta de suporte, especialmente considerando que parte significativa dos nossos membros é composta por pessoas idosas, que podem apresentar maior dificuldade na navegação pelo fluxo automatizado.Como sugestão de melhoria, entendemos que seria importante revisar a lógica de encerramento automático e criar mecanismos de fallback que direcionem mensagens livres para atendimento humano ou para uma nova tentativa de triagem antes do encerramento definitivo da conversa. Ou a opção de desligar a função
Already have an account? Login
No account yet? Create an account
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.
