Now Assist AI Agents Version Mismatch Between Dev and Test

[Sahithi K]

Hi Community,

Looking for guidance/best practices around Store app version mismatches across ServiceNow environments.

We have a scenario where a Now Assist-related Store application is on different versions between dev and test environments. The higher environment is already on a newer major/minor version, while dev is still on an older version.

My concern is:

• If we upgrade the app in dev to match test, could it change the behavior of existing AI agent capabilities/workflows?

• For example, behaviors related to generated plans, orchestration, prompts, or skill execution.

• If behavior changes in dev after the upgrade, how do teams safely validate and avoid issues when promoting future customizations to higher environments?

Would like to understand:

• Recommended approach for handling Store app version alignment across environments

• How teams test AI Agent/Now Assist behavioral changes after upgrades

• Whether app version mismatches themselves are a bigger risk than upgrading

• Best rollback or mitigation strategies if AI behavior changes unexpectedly

Any real-world experience or best practices would be appreciated.

Thanks!

[Dr Atul G- LNG]

Hi @Sahithi K 

When we upgrade and log in, the plugin fetches and updates the latest files and fixes for that particular module. If there are no changes in a few files, they may not be impacted by the plugin update.

In your case, you need to read the plugin release notes to understand what is included in the latest version. It is also recommended to keep all instances on the same plugin version to avoid issues like the one you are facing.

[Sahithi K]

Thanks for the clarification.

That helps. My concern was whether upgrading the Now Assist AI Agents app in dev could unintentionally change existing AI Agent behaviors even if our custom configurations remain the same.

I’ll go through the release notes in detail and validate the impacted use cases after the upgrade. Keeping all environments aligned definitely seems like the safer approach here.

Appreciate your input.

[Tanushree Maiti]

Hi @Sahithi K 

I’m just curious how it’s possible that, without any upgrade (perhaps even a patch version) in the sub-production environments (Dev, Test, and Prod-Stage) ,the Production instance gets updated directly!!

Upgrading a ServiceNow Production instance directly with major/minor version—without first upgrading sub-production (Dev, Test, Stage) instances—is highly discouraged.

While patches are often minor, bypassing the development cycle creates significant risks.

Probable Risks:

• Production Outage/Bugs: Unforeseen bugs in new code can break business-critical workflows, UI policies, or client scripts immediately.
• Broken Integrations: Upgrades can change APIs or data structures, leading to failures in integrations with third-party systems.
• Severe Performance Degradation: Performance issues introduced by new code can directly affect users, with no testing environment to diagnose them beforehand.

• Unmanaged Skipped Updates: When upgrading Servicenow skips records that were customized, marking them in a "Skipped" list. Without a sub-prod upgrade, these are not reviewed, meaning customizations might revert to OOB or break.
• Clone Discrepancies: If you clone Production back to Sub-Prod after the direct Prod upgrade, you are pushing a newer, potentially unstable version to Development. This makes future development impossible until Sub-Prod is also upgraded.
• Lost Work: If an incident occurs, you may be forced to clone from a backup, which could wipe out configuration changes ("in-progress" update sets) developed since the last backup.

 Mitigation Steps

Take these steps immediately to reduce impact:

• Clone Promptly: Clone the now-upgraded Production instance down to Development and QA immediately to align environments.
• Check Hi Support: Review any specific hotfixes or patches released by ServiceNow

Once your instance version is controlled, then you should starts with NowAssist plugin installation.

[Sahithi K]

Thanks for the detailed explanation. My concern was mainly around the AI Agent behavior changes after upgrading the Store app version, especially for capabilities like generated resolution/change plans and orchestration flows.

Test is already on the newer version, so I was trying to understand whether keeping dev on the older version would create more deployment and testing inconsistencies going forward.

Your points about skipped updates, environment alignment, and risks with mismatched versions make sense. I’ll review the release notes carefully and plan proper regression testing in dev before promoting any further changes.

Appreciate the guidance.