- Post History
- Subscribe to RSS Feed
- Mark as New
- Mark as Read
- Bookmark
- Subscribe
- Printer Friendly Page
- Report Inappropriate Content
2 hours ago - edited 2 hours ago
What's changing?
Beginning July 9, 2026, ServiceNow will begin rolling out third-party model providers as the default for out-of-the-box (OOTB) Now Assist skills and AI agents. The rollout is tied to individual store app release schedules, different products will make the switch at different times as their updates are released. If you've already configured a model provider, your configuration is preserved. Most admins won't need to take any action.
If you are a ServiceNow Isolated Markets or Self-Hosted/On-Prem customer, please contact your account team for more information.
In this article:
✓ Why ServiceNow is making this change
✓ What changes and when
✓ What you need to do — by scenario
✓ What stays the same
✓ Customers in air-gapped, self-hosted, and regulated environments
✓ Where to go for more detail
The AI landscape moves fast. Frontier models from Azure OpenAI, AWS Claude, and Google Gemini release capability improvements in reasoning, language understanding, and task completion, at a pace that directly benefits your Now Assist skills and AI agents.
ServiceNow manages that model lifecycle for you. When a new model version becomes available, ServiceNow evaluates it against OOTB skills and agents before making it the default, so your teams get the benefit of continuous improvement without having to manage the underlying infrastructure or model versions themselves.
By making these providers the default, ServiceNow gives you access to that cadence of improvement as a standard part of the platform, not something you have to configure or pursue separately.
Validated before the switch
ServiceNow product teams test and validate OOTB skills against the assigned third-party provider before any default changes. The provider for each skill is selected based on validated performance against that skill's use case, not an arbitrary assignment.
There are two separate events to be aware of. They're related, but they happen on different timelines and affect different configurations.
Starting July 9, OOTB Now Assist skills and AI agents where the default model provider hasn't been changed will switch to a third-party provider when you apply your product's store app updates. This rollout is tied to individual store app release schedules, so different Now Assist products will make the switch at different times. July 9 is when the rollout begins, not a single simultaneous cut-over across all products.
To see which provider has been assigned to your skills and agents after applying a store app update, check the AI Control Tower Impact Summary under Controls > AI Model Providers or the Now Assist Admin Console > Settings > Manage Model Providers.
Third-party providers aren't new
Azure OpenAI, AWS Claude, and Google Gemini have been available and fully supported on the ServiceNow AI Platform for several releases. What's changing on July 9 is that OOTB defaults will switch automatically when you take the store app update, not that third-party providers are new.
Now LLM Service general-purpose models will be deprecated and retired for commercial deployments on a separate timeline. If you've deliberately configured Now LLM Service as the provider for specific skills or agents, your configuration won't be overwritten by the July 9 store app updates. The forcing function for those configurations comes when Now LLM Service models are retired.
For current deprecation and retirement target dates, refer to the model status table in the Now Assist Model Lifecycle KB article →
Start here — regardless of your configuration
Before taking any store app update, review your allowed model providers in AI Control Tower. By default all providers are available, but if your organization has restricted specific providers, skills and agents assigned to a restricted provider won't function after the switch. For example, if your instance restricts AWS Claude and the new OOTB default for a skill is AWS Claude, that skill will be blocked by your own policy if fallback is disabled until you update your allowed providers or change the assigned provider.
Review your allowed providers under Configurations > Controls > AI Model Providers in AI Control Tower. Use the impact summary to see which skills and agents would be affected before you update.
Once you've reviewed your allowed providers, find your configuration scenario below.
An instance-level configuration applies to all Now Assist skills and agents on your instance, so your OOTB defaults are already fully covered. The July 9 store app updates won't overwrite that configuration.
If you've set the Now LLM Service as your instance level model provider, we recommend starting the process of getting at least one third-party model provider approved in your organization, performing testing, and setting your instance level model provider to the approved third-party one prior to the Now LLM Service retirement date.
Configurations set at the skill, skill group, or agent level are preserved when you take the store app updates. However, any skills or agents that don't have an explicit configuration, and are still running OOTB default of Now LLM Service will switch to the new third-party defaults when you update.
If any of your explicitly configured skills, skill groups, or agents are set to Now LLM Service, we recommend starting the process of getting at least one third-party model provider approved in your organization, performing testing, and updating those configurations to an approved third-party provider prior to the Now LLM Service retirement date.
You can review the OOTB defaults that have changed (not explicitly set) in Now Assist Admin Console under Settings > Manage AI Models > Manage Model Providers. The Model Providers tab shows provider assignments at the skill and skill group level, so you can identify which skills changed providers after you performed the store app updates in sub-production.
AI Stewards can also review default changes in AI Control Tower > Controls > AI Model Providers > Impact Summary.
For skills or agents that will switch, confirm the new OOTB default provider is on your allowed list in AI Control Tower. If it isn't, either add it to your allowed providers or set an explicit allowed third-party provider for those skills.
All OOTB skills and agents will switch to a third-party provider when you apply each product's store app update. No action is required to make the switch happen, after you update in sub-production, confirm that the new OOTB default provider for each skill is on your allowed list in AI Control Tower, or configure the skills to use an allowed third-party model provider.
After applying the update, verify the assigned providers in AI Control Tower under Configurations > Controls > AI Model Providers, or in Now Assist Admin Console under Settings > Manage AI Models > Manage Model Providers.
Explicitly configured providers are preserved — the July 9 store app updates won't overwrite them. However, Now LLM Service models will be deprecated and retired for commercial deployments. Now is the right time to begin planning your migration to a supported third-party provider.
Review your Now LLM Service configurations in Now Assist Admin Console under Settings > Manage AI Models > Manage Model Providers to understand the scope of what needs to move. ServiceNow will notify you ahead of retirement dates, but starting your migration planning early gives you more time to retest custom skills and validate prompt performance against the new provider.
For current deprecation and retirement target dates, see the model status table in the Now Assist Model Lifecycle KB article →
Custom skill configurations are preserved by the July 9 store app updates, nothing is overwritten automatically. However, Now LLM Service models are scheduled to retired for commercial deployments in December 2026, and prompts built for Now LLM Service models may not perform the same way on a different provider. Waiting until the retirement date to start testing leaves little time to address regressions.
Start now by reviewing your custom skills in Now Assist Skill Kit and identifying which are built on Now LLM Service. For each, clone the prompt and configure it to use a third-party provider, then run human and automated evaluations to compare output quality before the retirement date. This gives you a validated replacement prompt ready to publish before the switch is forced.
Don't wait for the retirement notification to start
Prompt tuning across a different model provider takes time. Check the Now LLM Service retirement target dates in the model status table, then work backwards to allow enough time for evaluation and iteration before that date. For step-by-step guidance on updating prompts for custom skills, running evaluations, and publishing updated skills, see the Now Assist Model Lifecycle KB article →
- AI Control Tower — AI Model Providers →
Review allowed providers, view the impact summary, and manage provider policies across your instance. - Now Assist Admin Console — Select a model provider →
Configure model providers at the instance, skill group, and skill level. Review the Model Providers tab for current assignments across skills and skill groups. - Exploring large language models on the ServiceNow AI Platform →
Overview of available model providers, model types, and how provider selection works across the platform.
The governance and data protections you rely on today apply regardless of which model provider is powering your skills and agents.
- AI Control Tower continues to give you visibility into model usage, performance, and policy enforcement across your instance, including which providers are active and where.
- Data protections are maintained by ServiceNow across all integrated providers. You don't need separate contracts with Microsoft Azure OpenAI, AWS, or Google to use these providers through the ServiceNow AI Platform.
- Your existing out-of-the-box skills and workflows don't need to be rebuilt. You don't need to re-engineer out-of-the-box prompts or modify agentic workflows to make this transition.
- Assist utilization and billing are unaffected. The new default models consume Assists at the same rate as the previous defaults.
- Your ability to configure providers remains exactly as it is today, at the instance, skill group, or skill level through Now Assist Admin Console, with AI Control Tower governing what's allowed.
This article describes changes for commercial deployments where connectivity to ServiceNow-integrated third-party providers is available. If your deployment falls into one of the following categories, this transition doesn't apply to you at this time.
- Air-gapped or isolated environments — where third-party provider connectivity isn't available
- Self-hosted deployments — where third-party provider access is not part of the deployment model
- Regulated or sovereign markets — where data residency requirements limit third-party provider availability
What to expect
You can continue using Now LLM Service for your generative AI skills and agents. ServiceNow is developing an open-weights hosted model package for these deployment types and will communicate availability and migration guidance separately when it's ready. No action is required at this time.
Your operational reference for everything related to model versions, upgrades, and retirements. It covers:
- The current model status table: all providers, models, release dates, deprecation targets, and retirement targets
- How to check which model your skills and agents are currently using
- How to access upgrade models before a scheduled retirement date
- How to validate that your instance has been upgraded successfully
- Step-by-step guidance for retesting and re-tuning custom skills after a model switch
- What to do if you notice a quality change after an upgrade
