Virtual Agent

Remote user's video rendering is intermittent on iOS devices when using @zoom/react-native-videosdk. The video feed from participants randomly works - sometimes it displays correctly, sometimes it fails to render at all. This issue does not occur on Android devices where remote video renders consistently.To ReproduceSteps to reproduce the behavior:Initialize Zoom Video SDK session on an iOS deviceJoin a Zoom session using zoom.joinSession()Have a remote user join the same session (from web or mobile)Render remote user's video using ZoomView component with their userIdObserve that remote user's video feed intermittently fails to renderExpected behaviorThe ZoomView component should consistently render remote participants' video streams on iOS, similar to the reliable behavior on Android. Once a remote user joins and their video is enabled, their video feed should display in the ZoomViewcomponent.Screenshotshttps://github.com/user-attachments/assets/dd8171b5-d8ed-4dda-8544-a93d22c83527htt

 ⚫️ Process to test AIC 3.0 and transition your current agent to AIC 3.0:Duplicate your existing agent Ensure the ZVA 3.0 toggle is enabled on the cloned agent Run test scenarios in preview to confirm expected behavior (Optional) Once validated, go to your production agent and scroll down to the bottom of the page to find the traffic splitting setting (Optional) Start by routing 10% of traffic to the cloned agent with ZVA 3.0 enabled, 90% of traffic to your existing agent. (Optional) Monitor containment rate, resolution rate, and surveys to identify any adjustments needed in the ZVA 3.0 agent. Make adjustments and publish the AIC 3.0 agent. Gradually increase the traffic routing until you are ready to enable ZVA 3.0 on your production agent. Once you are at 100% of traffic on the AIC 3.0 agent, then you can go back to the AIC 2.0 original version, turn on AIC 3.0 and 100% of traffic on the slider. Delete the test agent.  ⚫️ Why did we launch AIC 3.0 or what is different? Read this: htt

Hello Support community team; I escalated ticket #TS2221540. But this ticket is only Trouble shooting. From Support agent, my request should ask to this community.Because this is ZVA’s development or make call center by using ZVA’s voice agent. I think the following 5 step's Scenario.(Now Widget menu is Blank) 1. Customer call test help desk#(050-xxxx-xxxx).2.If customer wants to Japanese, Push 1.If customer wants to in English , Push 2.3.Hearing customer situation.4.If need, ZVA's Voice agent ask to customer for catch up detail.5.If ZVA's agent understands, he or she suggests accurate solution.6.If above's suggestion's does not fixed, Voice agent ask to customer name and mobile number.7.Can we receive from step 3 to 6 conversation's story by e-mail?8. Form improvement of speech accuracy, I'd like to manage knowledge base, add too, and Intent function.(But I'm not sure their steps)Would you mind sharing above step by step in detail by April 2nd?(I'm beginner ZVA)I appreciate your dedic

The Arch Linux tarball file downloads. However the build or just tar xJf fails with an “unexpected EOF in archive”.Here is example output of the build and the use of tar command.Package (1) New Version Net Changezoom 7.0.0-1 935.63 MiBTotal Installed Size: 935.63 MiB:: Proceed with installation? [Y/n] y(1/1) checking keys in keyring [------------------------------------] 100%(1/1) checking package integrity [------------------------------------] 100%(1/1) loading package files [------------------------------------] 100%(1/1) checking for file conflicts [------------------------------------] 100%error: failed to commit transaction (conflicting files)zoom: /usr/bin/zoom exists in filesystemzoom: /usr/share/applications/Zoom.desktop exists in filesystemzoom: /usr/share/mime/packages/zoom.xml exists in filesystemzoom: /usr/share/pixma

Ever tried to prompt and get the LLM to gather details from the customer and then use that to make decisions? Has it ever failed? That’s because your architecture for how variables and LLMs interact might need a refresh. Best practice 1: Always use a tool to set the variable and then actually map it to a real global custom variable. Yes, it takes longer in the beginning but it’s worth it for the consistency in experience. Best practice 2: Create a set of 10 “backup” custom variables for you to use as placeholders while you build. You can then come back and rename them to the correct value they represent or exchange them with a new variable. This helps keep you in the habit of actually mapping to a variable field, saves time while you build, and let’s you be more creative with your UX design. Keep a list of what you used the “backup” variables for and always go back and rename them or make new variables to switch out. Best practice 3: Don’t just prompt a reference to a variable. Actuall

Hi everyone,We’ve been using Zoom ZVA’s built-in transfer_to_live_agent tool to route customers into our ZCC flow. Overall it works well (around 90–95% success rate), but we’re running into an unavoidable issue where the transfer fails because the required variables aren’t being set by the agent properly before execution.  These variables are critical on our end since they determine which queue the customer gets routed to.Has anyone else experienced something similar with variable handling during transfer? If so, how did you resolve it or work around it?Any insights would be amazing!Thanks,Arpit 

Has anyone been able to actually reach a human for support? The virtual agent gave me a ‘supervisor’ which sounds like a fake title given out and then that person gave me an email that bounces back email. I can see why they are loosing customers.

You’ve seen ​@ShawnFerrell’s awesome instructions on how to build AI Receptionist (concierge) for Zoom Phone…..Now see it in action as I challenged myself to build one in under 10 minutes!Zoom have packed a huge amount into AI Receptionist & made it really easy to deploy. Check out the video on YouTube & leave a comment on the channel for what you would want to build!  

Important Note: This guide is tailored for the ZVA 3.0 experience. If you’re using 1.0 or 2.0, please let me know in the comments and I’ll happily do a write-up on how to build an AI Receptionist for your ZVA version!Why Build an AI Auto-Receptionist?If you manage Zoom Phone, you already know the Auto Receptionist. It’s the front door of your phone system — callers dial in, listen to menu options, press a number, and get routed. It works. It’s reliable. It tells you almost nothing about why people are actually calling. Zoom Virtual Agent (ZVA) changes that. Instead of routing calls based on key presses alone, a virtual agent listens to what callers say, understands their intent, and routes them accordingly — while capturing data on every interaction. It even assists with deflection, by being empowered to handle FAQs that bog down your frontline teams. The result is the same routing your team already manages, plus analytics that reveal what your callers actually need, enabling you to ta

Hey Folks!Depressingly this video is already out of date as 3.0 very recently launched, but hopefully it will still help you familiarise yourself with ZVA build & moving away from the classic chatbot 1.0 concepts.https://youtu.be/3PjVg2hfq6k

If you are asking a series of questions to collect information but if the customer’s answer then determines the next question (an “if-then-else” prompt) this is the best practice with ZVA Agents: 1. Turn on AIC 3.02. Make an instruction skill 3. Limit “if-then-else” prompts to five scenearios using an instruction skill. If you use  six or more “if-then-else” statements in your prompt, then the experience will degrade because the context of the conversation is not best matched for one agentic conversation. If you have more than five “if-then-else” statements in a single prompt, use these questions to continue following best practices: » Why do we need these questions? » Can we include the intent of these questions inside of questions #1-5? » Will my customer be fatigued at this point and give false or incomplete answers in order to “just get it done?” » Have you prepared the caller before starting the “if-then-else” statements that this process take more than 5 minutes to complete? To c

I’ve been playing with ZCA quite a lot recently & just added the latest video to the #apieceofCX library with a look at ZVA Analytics.Very impressive & super accessible.Watch on YouTube here!

 Check out the features from January that we recommend you consider for your ZVA management practice! All features listed here: https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0067812

Hi everyone. I'm trying to setup the integration between AVZ and Manage Engine ServiceDesk Plus.Based on the KB https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0057946 I insert "itdesk" and select my region (EU). I'm redirected to the auth page on Zoho but after the authentication process I land on an error page that says:  Invalid ClientClient ID passed does not exist Any idea?Is the ServiceDeskPlus supported or do I miss something?"itdesk" refers to other ManageEngine products? RegardsLuca  

Hello, I need your helps. I got 050 phone number by eKYC and paid “Charge Name:Zoom ContactCenter Monthly Usage -”. But I could’t find voice agent menu(I showed chat agent and classical chat bot menu only).I’d like to make the simple IVR(Japanese or English and Loss the urgent matter(mobile , PC) or Admission Application) using Zoom Virtual Agent simply.Would you mind sharing the step by step the flow guide immediately?Also I’m very happy to the guide with Japanese directly. I appreciate your dedicated support for Japan in advance, Hiro.  

I would like to be able to filter matched queries that I have already coached and trained. I would expect coached queries to drop from the list of matched queries or there should be a filter to hide "coached queries". Do they never disappear? If not, how is that scalable? If they should be dropping from the list, what step am I missing? My steps are:1. Navigate to Intent Management.2. Scroll to an Intent group.3. Select an intent within the group.4. Click on the Coach tab.5. Click the check box next to queries to be coached.6. Add to training phrases.7. Train.8. Publish. The use case for why it is problematic is sorting through multiple identical or similar queries to find the queries that I have not coached. Below is a real-world example of one of my intents:1. I have an intent for Financial Input.2. I have 195 queries for "Financial review," 4 queries for "Financial input," and 1 query for "Can I get financial input, please?"When I have 5000 queries, how will I filter out "Finan

We have a complex contact center flow that is working well but we cannot stop the transfer tool from saying its inner chain of thought inspite of various guardrails.Any ideas? None of our prompts provide instructions on reading content back to the caller.eg. AI Agent Hello. How can I help you today?Consumer How can I set up a Saturday morning detention?AI Agent Okay. Let me transfer you to one of our support agents who can help with this. Can you please tell me any more details we may need to know before I transfer you?Consumer Uh, our period structure is just a Monday to Friday.AI Agent Right. Thank you for that detail. Is there anything else you'd like to add before I transfer you?Consumer No.AI Agent Setting up a Saturday morning detention. The period structure is currently Monday to Friday only. No further details provided.AI Agent Here's what I've captured: Wants to set up a Saturday morning detention but period structure is Monday to Friday only. Connecting you to an agent now.AI

Make sure that you are optimizing your agent for the job they were “hired” for. Do not prompt-bomb the guidance and write a Da Vinci Code length guidance prompt. It’s hard for a human to read and hard for AI to understand. So, what is the ideal length of your guidance? First, factor in that ZVA is not only digesting your guidance prompt but also, Length of the conversation - which we are optimizing using summary. Number of skills and their complexity. How tool responses are being returned. How guidance is written as well (the architecture and formatting matter).  Best Practice:Target for one agent’s guidance prompt: 2,000–4,000 characters(~500–1,000 tokens)Why this range works bestEnough room for:role definitiongoalstoneguardrailstool-calling rulesShort enough that the model consistently attends to itEasier to debug when behavior goes sideways Upper limit: 8,000 characters. You won’t be stopped, but you also won’t see the results you nor your users want. Once it goes past this, you’re

I am developing a custom solution using Zoom Virtual Agent (ZVA) where customers provide image inputs during chat interactions. However, ZVA does not support image uploads out of the box, as confirmed in my previous case with you. In the CX Analytics tab > Logs > Transcripts, I can see the chat history, including attached images , which appear in the UI. To work around this, I want to programmatically fetch the transcript and the image attachment using APIs, as text responses can be passed via variables, but images require API access for processing (e.g., in AWS Lambda). I have explored the following APIs with a Server-to-Server OAuth app (scopes include contact_center:read:engagement:admin and contact_center:read:inbox_messages:admin): GET /contact_center/engagements/{engagementId}/events Response: {"code": 7001, "message": "Only voice and video channels are currently supported."} Limitation: Does not support messaging channels. GET /contact_center/messaging/messages?engagement_

hi Zoomies.  I didn't see a topic area for AIEA, so i am posting here in ZVA.  I am setting up a simple chat flow and queue as a customer entry point for general inquires.  The queue will have the Expert Assist enabled (KB and NBA).   Since you can only have one AIEA instance assigned to a queue.   AND, since each AIEA can only have one Intent Group. (I think) Then, if your queue is for general inquiries,  won't your Intent Group have to be very broad (and possibly large). I'm hoping someone can tell me I'm missing something.   In a perfect world, it would be nice if I could assign multiple Intent Groups. 

Hello, My company has signed up to Zoom Contact and Zoom Phone Enterprise. I have a good working knowledge of AI voice agents having worked on POC agents in Ragflow, N8N and Elevenlabs. It's early days so I am happy to park this to one side as we are working with a professional partner.What I am struggling with is the change in behaviour between querying a knowledge base, querying a virtual agent with a minimal prompt and querying a  deployed virtual agent in full screen url campaign deployment. Given I have spent much of the last year, preparing, cleaning and building RAG agent  workflows for text and speech, I am trying to figure out why the same knowledgebase would retrieve different matched content from the same knowledge base using a vanilla prompt.Are there video tutorials and documentation that go through end to end example projects to give blueprints and examples for start to finish workflows, as is often the case with N8N youtubers?

When the Zoom Virtual agent submits a response, it plays a tone at maximum volume, which hurts my ears. The tone does not respect the Windows volume control. Environment: Dell Latitude 5400 laptop, Windows 11 Pro, current Zoom version.

If your knowledge base sync is stalling or creating an error due to size, please read this article. https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0058272If you're KB is under the size limit but still stalling when syncing, please submit a support ticket under the product = zoom virtual agent. On Monday mornings, the KB crawlers are slowest because clients like to update their KB's then. Try an off-rush hour time like Monday night or Tuesdays for a faster weekly sync. 

We are preparing for the transition to Salesforce Enhanced Messaging from Live Chat. As a company with support agents around the clock and around the world, we rely heavily on a script to check agent availability in Live Chat. As Salesforce will not be providing a similar solution for Enhanced Messaging, I'm curious how other companies are addressing this critical workflow. Suggestions?