AI Assistant

Hello Moveworks Community,We are reviewing our Triage 2.0 prediction model threshold and would like to learn from other customers.Could you share:What model threshold are you using? What precision and coverage are you seeing at that threshold? Why did you choose that threshold? Did you test other thresholds before selecting it? How long did you monitor the results before keeping the threshold? What would cause you to change the threshold?Please state whether your precision and coverage results come from Production, a model test, or a training report. This will help us compare similar results.We understand that each customer has different ticket data and routing needs. We are interested in how other customers evaluated the balance between precision and coverage.Thank you.

Hey all,I’ve been having the following error all morning in the Agent Architect:“An Error Occuredexception TypeError: fetch failed sending requestDetailsexception TypeError: fetch failed sending request”The agent gets stuck on “Analyzing your request...” for a few minutes, then spits out the same error. It happens almost every time I prompt the agent. When I refresh the page, the error and prompt are completely gone. I have to repeat the prompt 2-3 times for the agent to actually think and give me a response.Any ideas how to fix this? It is really slowing me down when trying to build out new plugins.Thanks!  

We are currently unable to send a communication in the language by default, or even just to offer the recepient of a campaign the option to read the campaign in French or English. This is causing multiple issues, wether we wish to send surveys, or longer comms, as we always have to send a single comm bilingual (French on top, English. at bottom. Hopefully this will be addressed soon as your company grows larger (especially sinc being acquired by SNow),

We send notifications to users through Moveworks (ambient agent) and surface them in the assistant inside ServiceNow. Today a user has to open the notifications bell and finds the right one. Is there a way to get a direct link to a specific notification's conversation so we could email someone a link that opens that exact notification in the assistant, instead of making them dig through the bell? If so, is that info available right away when the notification is sent, or only later?

Hi MW Community,New to MoveWorks and I want to understand if a use case I am exploring is possible. A simple use case example:Open the assistant with the context that my favorite color is blue. This will need to happen behind the scenes without me explicitly telling it. Assistant able to answer ‘what is my favorite color’ by referencing the context.Some open questions:  1. Does the Moveworks embed/web SDK accept arbitrary context on init that becomes available to the reasoning engine?                                                                                              2. Is there a "conversation starter" or "welcome plugin" concept that fires automatically on open?                               3. Is there a way to initiate a plugin to read from URL query parameters on open conversation? Thank You!

Was really excited about ‘Vetted Content’ being released (https://help.moveworks.com/ai-assistant/enterprise-search/vetted-content) however even after adding a few articles across different systems (confluence - next gen connector and google drive - regular connector) docs are showing as Available but not highlighted as vetted sources.The knowledge returned is correct, and its returning the correct knowledge articles, but based on “What end users see” https://help.moveworks.com/ai-assistant/enterprise-search/vetted-content#what-end-users-see it should be flagged. Seeing similar behavior between the slack and web assistant interface. Opened a support ticket, but figured it would be faster to poll the community than get an answer from support. 

Hey,I am reaching out to check if anyone has had success with the GitHub ingestion setup outlined in this documentation: https://docs.moveworks.com/ai-assistant/enterprise-search/configure-search/ingestion/how-to-configure-github-knowledge-ingestion.I configured the connector and ingestion as per the article. The records are successfully being ingested from the repo, but the base URL is being dropped and the .md suffix is being stripped, resulting in broken links (System doesn’t allow me to leave “Github Config Article Url Filepath Suffixes To Drop” as blank).For example, src/ad/README.md redirects to https://src/ad/README. The full path should actually be: https://github.com/newrelic/opentelemetry-demo/blob/main/src/ad/README.mdFor context, my ingestion config Start URL is: https://api.github.com/repos/newrelic/opentelemetry-demo/contents/Am I missing a step in the configuration to retain the base URL and the file extension? 

Hi Moveworks,Back in January, the Microsoft partnership with Copilot was announced. Where can I find more information on this integration? We just launched Microsoft Copilot for our org and we would like users to have access to our Moveworks Assistant in this new environment. 

I need to change the placeholder text displayed in the search field of the Moveworks Assistant.Currently, the text appears as:“Find answers, automate tasks, or create content instantly.”I would like to update this placeholder text.Could you please confirm whether this change is configurable? If so, kindly provide the steps or location within the Moveworks setup where this modification can be made. 

In the analytics that Moveworks provides, all knowledge that is used by the AI assistant is pooled into one category - Enterprise Search / Knowledge Base. I am unable to differentiate between which knowledge source information is being pulled from unless I go into each individual interaction. This makes it difficult to see performance of each source and also difficult to identify gaps or redundancy. The raw interaction table will show that Enterprise Search plugin was both successful and unsuccessful which can make things confusing. Is anyone else running into this issue? Are there any tips and tricks I could be missing?

Despite clear user utterances (e.g. payslip, salary, tax form, leave balance), the  Moveworks AI often fails to classify topics correctly, pushing high‑volume demand into “uncategorized” buckets. Improving intent‑to‑topic mapping here would increase accuracy and , insights quality. How can it be done better? Or is there anything on our side that we should consider improving? We also noticed that Moveworks Analytics struggles to categorize a  Domain when citing mainly (only?) pdf files when generating response and it states Domain as “N/A”. Is it known limitation?

I have a user ID in servicenow called as “suhas.shimpi” and email address as suhasdigamber.shimpi@mycompanyXXX.com. Now In moveworks i have manually created a user with same email address as above. I have super admin in moveworks. I have configured ITSM ID Info as well with Below details:ITSM User ID = suhas.shimpi (same as servicenow user id)External ID = “SYS ID of my user record from ServicenowFirst Name = SuhasLast Name = ShimpiFull Name = Suhas ShimpiIntegration ID = <Servicenow build connecter name that we used while connecting Moveworks to Servicenow > Now with all above details, When i am trying to use Moveworks Assistant and askign to file a ticket, it comes back with response as below: Help me if i have misconfigured anything. Please note that right now i am not using Okta integration for authentication. I have just manually created users in both systems with same unique identifier (email address).

Anyone else having issue with AI Agent, has stopped working since this morning 29th Apr ESTAn Error Occured{"error":{"message":"{\n \"error\": {\n \"code\": 400,\n \"message\": \"Unable to submit request because thinking_level is not supported by this model.. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/gemini\",\n \"status\": \"INVALID_ARGUMENT\"\n }\n}\n","code":400,"status":"Bad Request"}}

In Analytics, we review the Unsuccessful Knowledge plugin to locate gaps. We noticed that when a form is served, we received an “unsuccessful” for knowledge. Is anyone else seeing this? What are you doing to remedy this issue? 

  Getting the below error while creating new ticket configuration with ServiceNow connector and saving the configuration. There is no issue with the service account and SN connector.Failed to create integration: No bot service account found for this integration. Please configure a bot service account in identity settings for this integration. Error Code: 14128 If the error persists, please contact the support team with the following trace ID - F4z2BWmyKWu2peJ0

Is there a way to customize the information being presented in web assistant during the guided walk through users receive? Specifically the “Ask questions or automate tasks” box.See images below for what I am referring to. Some of the prompts are offering things to ask, that we are not yet connected to and we don’t want users to be misguided.  

Has anyone integrated with Navan (https://navan.com/)? We are hoping to be able to do travel booking and other tasks in Navan through our AI assistant.

Why is this happening? The agent should only be capturing the answers and citations, not the thinking steps (i.e. ‘Searching for’, ‘Summarizing based on’) 

Now that Assistant on Web has been live for all customers, we’re eager to understand how your rollout is going and what support you need as you introduce this new surface to all your employees.Since access to Assistant on Web is controlled through your organization’s My Moveworks SSO setup, we don’t have full visibility into how broadly it has been enabled across your organization. To help us improve the product and better support your rollout, we’re asking for 2–4 minutes of your time to answer a few questions via our survey. 🌟 Take our Survey We’d love to hear: How many employees you’ve enabled Assistant on Web for Whether you plan to roll it out more broadly What’s holding you back if you haven’t What would motivate you to expand access Any product gaps or change management needs Your insights will directly shape future improvements to Assistant on Web and how we help you activate it across your organization.

Hi,We have seen that late last year the Moveworks chatbot was not auto-scrolling to the beginning of the response in both the Moveworks web app, and the Embedded Assistant in ServiceNow. It seemed it was fixed briefly at the beginning of the year (end of Jan 2026 when MW branding changes were initially pushed), but now it seems like it is no doing the auto-scroll to the beginning of the response. Can we have this auto-scroll feature returned so that it works like it did at the beginning of the year in both the Moveworks Web App and in the ServiceNow Embedded Assistant?  Thank you,Deana

We are working on use case where we want the bot to respond to the users query of  filling the timesheet , the bot should respond with a table to fill the hours for the whole week , also a dropdown to select a date range like week or month , but as per my understanding the bot right now do not support the adaptive cards ? Am I right here ? Also I would like to know if anyone has built a similar kind of use case where bot prompts the end user with a table like structure to fill the data in and then we in the backend will be able to process it .  

I have a onboarding process that is initiated by a listener. The listener uses notify to contact the new hire’s manager. I need to collect information about what the new hire needs so notify has a button to start a conversational process. When this happens, the manager is asked for the information. However their response is initiating other plugins if there are some matching keywords. There will be some overlap here, so what are some suggestions from preventing the response from starting a new plugin?Note that this was working fine a month or so ago. Since then many new plugins were made which might be conflicting with this. I am leveraging many applicable positive and negative prompts in these new plugins.

Hi Community,I’d like to understand how your teams are handling ungrounded and hallucinated responses from the bot.Hallucinations are relatively easier to spot since they usually come without citations. However, ungrounded responses tend to create more confusion for end users. In most cases, users don’t read the full cited article and rely mainly on the summary provided by the bot, which can sometimes be misleading.For example, when a user asks to unlock their account, the bot surfaces an article about password reset and confirms that this will also unlock the account—which isn’t always true. This leads to incorrect expectations and user confusion.I’d be interested to learn:How are you detecting or mitigating ungrounded responses? Are there any best practices you’ve adopted to reduce user confusion in such cases?Looking forward to your insights and experiences.