When a ServiceNow AI agent is stuck, start by identifying the execution layer that stopped rather than repeatedly changing the prompt. Confirm that the Now Assist AI agents plugin, dependent applications, compatible instance release, configured agent and executing-user roles are present. Reproduce the run in AI Agent Studio, use the Decision Log and Activity replay, then trace the execution plan through its tasks, tool executions, messages and generative-AI logs. Cancel an execution plan that remains In Progress without activity, correct routing or tool metadata, reduce oversized inputs, add explicit fallbacks, and test ACLs, mandatory fields and data policies under the same user identity used in production.
This sequence covers the common messages There are no agents available at the moment, a tool that appears frozen, a Flow Designer operation failure, an employee task that never completes, and an agent that chooses the wrong tool. It also separates a genuine planner problem from a downstream record, permission or presentation failure.
As an Amazon Associate I earn from qualifying purchases.
What a “stuck” ServiceNow AI run actually means
An AI-agent request passes through several distinct layers. A trigger and user message are routed to an agent; the agent creates an execution plan; the planner invokes tools; tools read or write records; Flow Designer or another action commits the result; and a Now Assist panel or Virtual Agent presents a response. A run can look identical to an employee while failing at any one of those layers.
- Routing failure: no agent is considered proficient for the trigger, or an overlapping description sends the request to the wrong agent.
- Planner or tool failure: the selected agent loops, sends an oversized payload, receives an unusable response or waits on a tool that returns too many rows.
- Execution failure: the worker user lacks an ACL, cross-scope permission or required role, or a target record violates a mandatory field or data policy.
- Presentation failure: the downstream task changes correctly but the employee receives no usable completion message.
Do not treat a spinner as proof that the model is still thinking. The first execution record with an error or failed status is usually the most useful breakpoint.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
Check prerequisites before changing prompts
Confirm the required application and agent
The Flow Designer Use an AI agent action depends on the Now Assist AI agents plugin and on an agent that has already been created and configured. A missing plugin, disabled dependent application or unconfigured agent can produce a misleading automation error even when the flow itself is valid.
Record the release and application versions
Write down the instance release, the AI Agents Store application version and all dependent Now Assist plugin versions in the incident. Update those components according to your organization’s change process before redesigning prompts or tools. A version-specific defect can otherwise be mistaken for a routing or instruction problem.
Verify authoring and execution roles
Users who add the action need the flow_designer or admin role. The user who actually executes the flow may be different from the designer. For supervised execution, a support user must satisfy the AI security controls and be available to authorize the run. A designer’s access therefore does not prove that the production worker can read, create or update the same records.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Capture the production identity
Before reproducing the issue, record the channel, user, impersonation state, domain or scope and target record. Virtual Agent, Now Assist and a background flow can supply different identities or data contexts. Keep those values unchanged during the first diagnostic comparison.
Reproduce the failure in AI Agent Studio
- Open the agent in AI Agent Studio and use its testing page with the same trigger wording, record context and input that failed for the employee.
- Open the Decision Log for the run. Confirm which agent was selected, which tools were considered and which tool was actually called.
- Use the Activity tab to replay a prior execution. Save the run identifier and the timestamp so you can compare one change at a time.
- Keep the user, record, trigger and input constant while testing. Change only one prompt, description, filter or role at a time; otherwise a successful retry will not reveal which change mattered.
The Decision Log answers a routing question, while Activity replay gives you a stable execution to compare after each correction. If Studio succeeds but production fails, focus on identity, ACLs, channel context and downstream data rather than immediately rewriting the agent instructions.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
Trace the execution plan from the top down
Open the execution-plan record
ServiceNow stores conversations invoked for AI-agent execution in the Execution Plans table, sn_aia_execution_plan. Open the record for the failed run and inspect its status, timestamps, selected agent, trigger and plan details.
Follow the child records in order
Continue from the plan to its Execution Tasks, Messages and Tool Executions. Then inspect the generative-AI log metadata and message records when the visible task status does not explain the failure. The order matters: a planner can be healthy while a tool fails, and a tool can succeed while a later record update is rejected.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- Find the first task whose status changes to Error, Failed or another terminal failure state.
- Open the associated tool execution and compare its input payload with the tool’s declared input fields.
- Read the returned output, error text and row count. An empty or malformed output can cause the planner to repeat a call even when the tool technically completed.
- Match the timestamp to Flow Designer execution details, ACL checks and data-policy messages.
- Check the message record that should have been shown to the employee. A missing response after a successful write is a presentation or handoff defect, not necessarily an AI defect.
Cancel a plan that is genuinely stuck
If an execution plan remains In Progress with no new task, tool or message activity, locate it in the execution-plan list, cancel it and save the plan, task and tool identifiers before retrying. Do not cancel a plan merely because a slow external operation is expected; first confirm that timestamps and child records have stopped advancing. Repeated planner calls to the same task or tool are a sign of a loop, missing stop condition or unusable tool output.
Fix “There are no agents available at the moment”
This message normally indicates a routing or proficiency problem rather than a missing employee record. First update the AI Agents Store app and dependent Now Assist plugins if your release is behind. Then inspect the fields the orchestrator uses to decide proficiency:
- Agent name: make it specific to the capability and avoid names that differ only by a department abbreviation.
- Role or proficiency: state the records, operations and audience the agent is intended to handle.
- Description: describe the trigger, required context and boundaries in concrete terms.
- Trigger wording: use language employees actually submit, including the object and desired outcome.
- Overlap: remove descriptions that claim the same task for multiple agents unless a clear routing distinction exists.
Test a narrow request first, such as creating one defined employee task with a known record. Broaden the description only after that request routes consistently. If no agent is selected in the Decision Log, continue with names, descriptions, roles and trigger wording; changing a tool will not repair a routing decision that never reached the tool.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Stop the agent from choosing the wrong tool
Make tool metadata unambiguous
The planner relies heavily on tool names and descriptions. Give each tool a unique operational name that states the action and object, such as “Create onboarding task” rather than “Update record.” Describe required inputs, output fields, side effects and the next step that consumes the output. Avoid two tools whose descriptions both promise to update an employee task.
Free tools Windows power users keep installed
One-click scans. No signup required.
Describe an algorithm, not a broad goal
Instructions such as “help the employee with onboarding” leave too many valid tool choices. Specify the sequence: identify the employee, verify the onboarding case, collect the task type and due date, call the task-creation tool, validate the returned identifier, and present either a completion message or a safe handoff. Include a stop condition after a successful write.
Inspect the selected tool in the Decision Log
If the wrong tool is called, compare the user’s exact wording with the competing descriptions and required inputs. Rename or narrow the competing tool, state when it must not be used, and clarify how its output differs. Replay the same Activity record and verify that the planner now selects the intended tool before changing unrelated permissions.
Reduce tool load when a tool appears frozen
A tool can appear stuck when it receives too much data or must return too many rows. Start with the smallest useful record context and expand only when the run is stable.
- Add conditions that constrain the table, employee, case, date range or status.
- Limit the result set and return only fields the next step consumes.
- Filter before the tool call instead of asking the model to discard irrelevant rows.
- Inspect the tool execution payload and response size in the execution plan.
- Test an empty-result case and a multi-row case separately.
Large inputs also increase the chance that the planner loses the operational instruction among record data. Keep the prompt focused on the current decision and pass a compact, typed result from the tool.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
Add explicit recovery branches
Every tool should define what happens for an error, an empty result, a missing field and an ambiguous request. An interactive agent can ask the employee for the missing value or clarification. A non-interactive agent cannot ask a person, so its fallback must end in a clear failure message, a queued human handoff or another safe terminal state. Never leave the planner with an implied instruction to “try again” without a limit.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Resolve permissions, ACL and data-policy failures
Run the failing step as the production worker
Reproduce the failing action under the same worker or execution user used in production. Verify table and field ACLs, role assignments, cross-scope access and any impersonation differences. A successful Studio test under an administrator does not validate a production flow executed by a restricted worker.
Read the exact data-policy exception
Flow Designer can fail while inserting or updating an employee task because a data policy rejects the payload. A common example is a missing mandatory Assigned To value. Treat this as a data-contract failure: inspect the action inputs and mapped data pills, identify the required field named in the exception, and validate the target record manually with the same user.
Check field mappings before adding roles
Do not solve a missing value by granting broad permissions. Confirm that the previous tool returns the field, that the data pill maps to the expected type and that the target table accepts the value in the current scope. If the field is intentionally unknown, route to a clarification or human handoff instead of writing a partial employee task.
Compare channel-specific identities
Now Assist, Virtual Agent and an asynchronous flow may use different identities, domains or session context. Compare the user and worker recorded in the execution plan with the identity used by the successful Studio test. This often explains why the same instruction works in one channel and fails in another.
Make employee tasks complete and visible
For HR service delivery use cases such as onboarding, policy questions, leave, benefits or task creation, success has two parts: the downstream record must reach the intended state, and the employee must receive a usable response. Check both.
- Confirm the task or case was created or updated exactly once.
- Verify assignment, due date, state transition and notifications.
- Check that the employee can see the resulting status in the channel they used.
- Confirm that the final message includes the task identifier or next action when appropriate.
- Test a missing-information path so the employee receives a question, failure message or handoff instead of waiting indefinitely.
When execution fails, the failure message should be presented in the Now Assist panel or Virtual Agent. Inspect message records if the backend error is visible to administrators but no response reaches the employee. A correct database update with no user-visible handoff is still an incomplete automation.
Common failure signatures and first actions
| Symptom | Likely layer | First action |
|---|---|---|
| “There are no agents available at the moment” | Agent routing or proficiency | Update the Store app and dependent plugins, then review agent name, role, description and trigger wording. |
| Correct agent in Studio, wrong behavior in production | Identity, permissions or channel | Compare worker-user roles, ACLs, record context and channel identity; inspect the production execution logs. |
| AI agent tool appears stuck | Input size or result volume | Add filters and limits, inspect the tool payload and response, and add an error or empty-result fallback. |
| Agent repeats the same action | Planner loop or unclear stop condition | Review the instruction algorithm, tool output shape and stop/fallback rules; look for repeated calls in the trace. |
| Flow fails while inserting or updating an employee task | Data policy or mandatory field | Read the exact policy exception and populate required fields such as Assigned To before retrying. |
| Flow pauses waiting for a person | Supervised execution | Confirm the support user is valid and authorized, then complete the approval in the Now Assist panel. |
| Employee sees no useful response after failure | Presentation or handoff | Inspect message records and verify the Now Assist or Virtual Agent failure-message path. |
Choose the least risky remediation
When several fixes look plausible, compare them before making a production change. A prompt or filter edit is usually more reversible than a role or schema change, while a permission change can affect every employee workflow that uses the same table.
Recommended Free Tools
| Remediation | Failure layer | Reversibility | Typical blast radius | What to observe next |
|---|---|---|---|---|
| Clarify agent or tool name and description | Routing | High | Requests matching that description | Decision Log selection and competing-agent overlap |
| Add a prompt step or stop condition | Planner | High | Runs using the agent | Number and order of tool calls |
| Add filters and result limits | Tool load | High | The affected tool invocation | Payload size, response rows and elapsed stages |
| Correct a data-pill mapping or mandatory field | Data contract | Medium | The flow action and its target records | Insert or update status and policy exceptions |
| Adjust a role or ACL | Permissions | Low | Every workflow using that permission | Worker-user access and audit records |
| Add a human handoff | Presentation and safety | High | Only error or ambiguity branches | Employee-visible message and queue assignment |
Prefer the smallest change that produces a clearer execution-log signal. Validate it in a non-production context, replay the same run, and then test a failure branch before expanding the change to other employee workflows.
Quick Recap
Prevent recurring failures
- Keep a runbook with the instance release, application versions, plugin status, agent and tool names, worker identity and known mandatory fields.
- Maintain one owner for each agent capability so overlapping descriptions are reviewed when a new agent is added.
- Document every tool’s inputs, output shape, row limit, side effects and fallback behavior.
- Include a terminal success condition and a maximum sensible retry path in every planner instruction.
- Test normal, empty-result, permission-denied, missing-field and ambiguous-request cases.
- Monitor execution plans for long-lived In Progress records and repeated task or tool calls.
- Verify that an employee-facing message is emitted for both success and failure, not only that a backend record changed.
A compact incident checklist
- Record the release, AI Agents Store app and dependent Now Assist versions.
- Confirm the Now Assist AI agents plugin, configured agent and required authoring roles.
- Capture the production worker, channel, record and trigger.
- Reproduce in AI Agent Studio and save the Decision Log and Activity run.
- Open
sn_aia_execution_planand follow tasks, tool executions, messages and generative-AI logs. - Identify the first error or failed status; cancel only a plan that is truly inactive.
- Correct routing, tool metadata, input size, filters or fallback branches as indicated by the trace.
- Run the same step as the production worker and resolve ACL, scope, mandatory-field and data-policy errors.
- Verify the employee task transition, assignment, notification and user-visible response.
- Replay success and failure paths after the change, then document the exact fix and scope.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
