Recently active topics

We've released the Clearer Outcomes for Tool Calls model upgrade to Basic customers. How are we improving the Assistant’s behavior?Failed or empty tool calls now show a clear failure or no-results state, along with guidance on what the user can do next, instead of appearing successful.When is the release date?Frontier: September 9, 2026 Standard: September 14, 2026 Basic: September 17, 2026Technical RestrictionsThis initial release primarily covers Agent Studio Plugins. MCP failure transparency, and more specific action-error messages are not included in this release and are being handled as separate follow-up work.DocumentationDocumentation describing reasoning and tool-failure behavior will be linked here when published.Read more about the release here.Learn more about our model upgrade program here.

Hello,As I saw mentioned previously, Build plugins are custom plugins that you create from scratch or customize to meet your specific needs . These require development work using Agent Studio. My team has identified some Workday use cases that do not fall within those listed as build plugins currently. Are those buildable plugins the only ones we can create and customize, or is it possible to create an entirely from scratch plugin? I’m asking generally speaking at this point. Thanks,  Todd

I am working on implementing/testing the Outlook plugins from the marketplace. They all seem to work individually so far but I seem to have an issue passing from one to another and can’t seem to figure out the issue. If I know a meeting I haven’t accepted and say I want to accept/deny that meeting directly it can handle that straight no issues.If I ask for a list of meetings I haven’t accepted it is able to provide a list but if I then ask to accept/deny a specific meeting from that list it doesn’t seem able to transition to the respond plugin properly. What would be the best way to troubleshoot the context issues here? The logs don’t seem to show the slots properly to determine what isn’t transferring over correctly.

In Service Now it is possible to structure knowledge articles in sections which you can collapse or expand / accordion format.There may for example be instructions which are written in the extendable sections, but we are not sure if the assistant can ingest and use that information.I was looking for official information on the Moveworks docs to cover the expanded / collapsed content topic but could not find yet.Do you have a reference to point me to, or can you confirm if there are any limitations regarding this?With the HTML-based articles we’ve seen that the information contained in the collapsible part is being found by the Assistant. Is this because it’s HTML or because extendable sections are readable?We were also reading into the nested content docs but not sure if from a Moveworks perspective accordion formatted content is considered nested content?An overall limitation guide when it comes to ingesting content in Moveworks would be great instead of needed to check specific topic

Hi all - Not sure if this behavior exists today, and if not maybe a future product idea? We currently have a few systems integrated into Moveworks for Enterprise Search Indexed Content, specifically Google Drive and Confluence, both using the max capacity connectors. With our Google Drive ingestion, we are specifying specific Shared Drives for ingestion. With our Confluence ingestion, we are ingesting all spaces. If someone were to embed/link to a Google Drive file from a confluence page, will moveworks follow that link and ingest the linked content? My thought here is that the Moveworks integration into Google Drive will likely have access to the file, but it may not be ingesting that specific shared drive, so I’m curious to know if it would ingest or not? If not, could we have this transitioned to a product idea? This seems like it would be a good solution to catch files that may be useful, but not ingested as they don’t live in an ingested shared drive. I understand that Google Driv

ToolSmith enters Controlled Availability on September 22, 2026, bringing a conversational way to build and maintain plugins in Tool Studio.Describe the workflow you want, review the proposed plan, and let ToolSmith help create and test the pieces behind it. You can also bring an existing plugin into the conversation to make a change or investigate a failure.What you can do with ToolSmithBuild a new plugin from a description. Explain the outcome and the systems involved. ToolSmith researches available APIs and existing assets, proposes a plan, and builds after you approve it. Edit an existing plugin. Describe what needs to change, review the proposed approach, and work through the update and testing in the same conversation. Debug a failing plugin. Provide the issue and relevant conversation context. ToolSmith can inspect execution logs, investigate the cause, and propose a fix.For example, start with: “Build a plugin that shows my open Linear issues.” Or bring an existing plugin and as

Ever wondered how many unused resources are quietly eating into your IT budget? It’s time to save resources before waste becomes business as usual. 💰Until now, tracking down orphaned cloud resources required a mountain of manual work. IT teams had to pull usage reports, match them to users, and send endless follow-up emails—only for those emails to get completely ignored while the bills kept piling up.With NAVI’s new Cloud PC Lifecycle Assistant, that process is now entirely automated.By pairing Nexthink with NAVI, we built a smart, proactive communication workflow that puts the decision directly into the user’s hands right through chat. Here is how the automated lifecycle works:📅 Day 21 of Inactivity: NAVI drops a friendly message checking if the Cloud PC is still needed. 📅 Day 25 of Inactivity: NAVI sends a 2nd reminder if no activity is detected. ⚠️ Day 29 of Inactivity: NAVI delivers a 3rd and final warning. ⛔ Day 30: If the user hasn't responded or logged in, the Cloud PC is au

 This is two days of real builds, worked side by side with the engineers and product teams who crafted the platform you're already building on. The same platform running over 100 billion workflows and 7 trillion transactions a year. Writing the code has never been the hard part. Shipping it is. Requirements, reviews, approvals, testing; that's where momentum stalls, every time. Fortunately, now[dev]26 is built around fixing that last mile. What these two days actually look likeDay one is an all-day hackathon on EmployeeWorks and Build Agent, with dev advocates, forward-deployed engineers, and sales engineers circulating the floor to help if you get stuck. Come with a team or find one when you get there: either way, you're building something real and testing the approach you haven't had time to try yet.  Day two opens with a demo-driven keynote that builds live on stage (no pre-baked demos here), then opens into open-choice labs across three focus areas. More coming soon on that! You do

Have you registered for now[dev]26 yet? This is two days of real builds, worked side by side with the engineers and product teams who crafted the platform you're already building on. The same platform running over 100 billion workflows and 7 trillion transactions a year. Writing the code has never been the hard part. Shipping it is. Requirements, reviews, approvals, testing; that's where momentum stalls, every time. Fortunately, now[dev]26 is built around fixing that last mile. What these two days actually look likeDay one is an all-day hackathon on EmployeeWorks and Build Agent, with dev advocates, forward-deployed engineers, and sales engineers circulating the floor to help if you get stuck. Come with a team or find one when you get there: either way, you're building something real and testing the approach you haven't had time to try yet.  Day two opens with a demo-driven keynote that builds live on stage (no pre-baked demos here), then opens into open-choice labs across three focus

 This is two days of real builds, worked side by side with the engineers and product teams who crafted the platform you're already building on. The same platform running over 100 billion workflows and 7 trillion transactions a year. Writing the code has never been the hard part. Shipping it is. Requirements, reviews, approvals, testing; that's where momentum stalls, every time. Fortunately, now[dev]26 is built around fixing that last mile. What these two days actually look likeDay one is an all-day hackathon on EmployeeWorks and Build Agent, with dev advocates, forward-deployed engineers, and sales engineers circulating the floor to help if you get stuck. Come with a team or find one when you get there: either way, you're building something real and testing the approach you haven't had time to try yet.  Day two opens with a demo-driven keynote that builds live on stage (no pre-baked demos here), then opens into open-choice labs across three focus areas. More coming soon on that! You do

How sort order works on the Data APIEach Data API endpoint returns records in a defined sort order. Because results are returned one page at a time, that order is fixed — it's what guarantees no record is skipped or returned twice as you page through results. An $orderby request won't change it, and because the call still returns successfully, a requested descending sort appears to be silently ignored rather than rejected.Endpoint Sort order        users last_updated_time, then id — ascending conversations, interactions, plugin-calls, plugin-resources created_time, then id — ascending      Our documentation currently doesn't mention this limitation. Hence we wanted to share this caveat. Workaround for getting the latest user recordUsers is already sorted by last_updated_time — just ascending. Your most recently updated records are at the end of the result set rather than the beginning, which gives you two straightforward options:- Page through to the last page. The final page

Hi Community!After an exciting first quarter, April is packed with major model upgrades, smarter search and a brand new developer homebase. We’re kicking things off with GPT-5.2 going GA, rolling out more powerful multilingual capabilities, and making it easier than ever to build in Agent Studio. Plus, don’t miss Knowledge26 in May and our lineup of April office hours to stay ahead of what’s next.📩 Save the Date: Knowledge May 5–7, 2026 | Las VegasDon't miss your chance to be part of Knowledge, where you and AI get to work.  Learn how Moveworks from ServiceNow is powering the AI front door. Learn more and sign up. 📹 April Webinar and Office HoursApril 2 — Developer Office Hours: Sign up now April 8 — Live Product Tour: Register now April 14 — Developer Office Hours: Sign up now April 30 — Developer Office Hours: Sign up now March Webinar Recap: AI Agents for Your Business Systems and How Motorola Solutions Scaled Enterprise AI Support with Moveworks 🚀 Product UpdatesCheck out the la

Salesforce Approval Succeeds, but Bot Receives success: false I’m working on a bot integration with Salesforce to allow users to approve or reject JCO requests directly from the bot.The overall flow is:Bot → HTTP Action → Salesforce → Approval/Rejection → Response back to BotThe interesting part is that the approval/rejection itself is working correctly. The action is successfully processed and the status is properly logged in Salesforce.However, when the same action is executed from the bot, the response captured by the bot shows:success: falseThis creates a problem because the bot uses the Salesforce response to determine the message shown to the user. As a result, the bot indicates that the action was unsuccessful, even though Salesforce has actually processed the approval/rejection successfully.What makes this more interesting is that when I test the same HTTP Action directly, using the same type of request, Salesforce returns:success: trueSo the behavior looks something like this:

Moveworks Approval Mirroring isnt functioning in Moveworks SandboxAlthough followed all steps mentioned in support document, unable to configure Approvals through Moveworks for ServiceNow. This is in Sandbox Env. 

Connecting the AI Assistant to your apps is now smoother. When a request needs access tosomething like Google Calendar, Zoom, or GitHub, the assistant handles it up front insteadof failing partway through.Pre-flight ChecksThe assistant identifies every connection a request needs before it runs, and asks for allof them at once. If a request needs both Google Calendar and Gmail, you approve both in asingle prompt, each with a clear description of what it allows.Auto-RestartOnce you approve, your original request continues automatically with all prior contextintact. No repeating yourself, no typing "retry." Step away mid-approval and you have a30 minute window to come back and pick up where you left off.What you need to doNothing. These features have been enabled by default across all orgs. No configuration or admin actionrequired. 

Credit Acceptance launched their AI Assistant with a deliberate rollout strategy. By intentionally naming their AI Assistant “Cara Beta,” they signaled that Cara would evolve and improve over time, which encouraged employees to engage and share feedback.At Cara's one-year mark, the team dropped the "Beta" label and relaunched with a new look and a batch of new capabilities based on user feedback. Adoption climbed from 40% to 60%.We sat down with Stephanie Wolanin, Business Process Optimization Manager, and Rupa Jakkula, Enterprise Cloud Applications Manager II, to talk through their rollout strategy, navigating change management, and how they drove adoption.What was Credit Acceptance aiming to achieve with the AI Assistant?Stephanie: The biggest thing was letting our service desk focus on higher-value work instead of simple things a user could self-serve but typically didn't, because the culture was to pick up the phone. We also really wanted to start adopting AI as an organization, an

Most teams launch their AI Assistant, see decent early numbers, then plateau. Sound familiar?We analyzed data across thousands of deployments and found a clear pattern: many organizations achieve good adoption but stall before reaching great adoption, because they treat it as a change management exercise rather than an organizational transformation.The top performers don't just launch and move on. Here's what they do differently.The two metrics that matterMAU% — breadth of adoption (average customer: 22%, top performers: 51%) WAU/MAU — depth of engagement, i.e. how often monthly users return weekly (average: 35%, top performers: 42%)Where you sit on these two dimensions tells you exactly what to focus on next.The three-stage playbookStage 1 — Build Your User Base If you're below 40% MAU, this is your only job. Awareness campaigns alone won't move the needle, you need to optimize the product and run a deliberate change management strategy in parallel. The three biggest levers: enable th

I am looking for guidance on response content. Does anyone have any suggestions or examples that they use. My concern is that our responses provide too much information. Here is a peek under the hood.  Utterance: Can you tell me how to set up a profile on LinkedIn?Can you tell me how to set up a profile on LinkedIn? Response:  

We've released the Clearer Outcomes for Tool Calls model upgrade to Standard customers. How are we improving the Assistant’s behavior?Failed or empty tool calls now show a clear failure or no-results state, along with guidance on what the user can do next, instead of appearing successful.When is the release date?Frontier: September 9, 2026 Standard: September 14, 2026 Basic: September 17, 2026Technical RestrictionsThis initial release primarily covers Agent Studio Plugins. MCP failure transparency, and more specific action-error messages are not included in this release and are being handled as separate follow-up work.DocumentationDocumentation describing reasoning and tool-failure behavior will be linked here when published. Read more about the release here.Learn more about our model upgrade program here.