--- sourceDocument: Australia Enable AI sourceDocumentLink: https://www.servicenow.com/docs/r/australia/intelligent-experiences Release : - australia ft:locale : - en-US ft:publication_title : - Australia Enable AI ft:clusterId : - platai bundleId : - platai workflow : - Platform --- # AI model providers # AI model providers {#ariaid-title1} Release version: Australia Updated August 14, 2026 ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 4 minutes to read Summarize ![AI sparkle icon](https://servicenow.com/docs/portal-asset/ai-sparkle-icon) Summarized using AI This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete. ## Summary of AI model providers AI model providers in ServiceNow allow administrators to configure data routing and manage access to third-party large language models (LLMs) and small language models (SLMs). The AI Control Tower offers settings to select which AI model providers are permitted on the instance, divided into: Show full answer Show less * **ServiceNow-supported providers:** such as Now LLM Service, AWS Claude, and Now LLM-LTS (Long Term Stable) designed for regulated industries with enhanced governance and compliance features. * **Organization-configured providers:** such as Perplexity and IBM Watson. ## Data routing and model management Data routing optimizes AI request traffic by directing LLM and SLM requests to the most suitable datacenter, reducing latency and improving response times. Two data routing types are available: * **Regional data routing:** Routes requests to datacenters within a specific geographic region, supporting compliance with regional data sovereignty requirements. * **Global data routing:** Routes requests to the best datacenter globally, regardless of the request's origin. Administrators can configure third-party LLM providers to use either routing type and specify allowed model providers. ## Model Preview Program (MPP) MPP is an opt-in program offering early access to AI models labeled as preview, beta, or experimental. Eligible users can test upcoming models, develop custom skills and agents, and provide feedback. Enabling MPP requires administrator acceptance of terms and toggling the MPP setting in AI Model Providers. ## Fallback and spillover features * **Fallback:** Allows active AI systems unsupported by enabled model providers to continue operating with their default providers. This is enabled by default but can be turned off, in which case unsupported AI systems must be deactivated. * **Spillover:** Addresses capacity limits in regional AI model deployments, preventing request rate limiting. Currently, this is active for Azure OpenAI when selected. ## Impact summary and support matrix The Impact Summary reflects the effects of allowed model providers and fallback status on AI system availability: * With fallback off, unsupported active AI systems must be deactivated. * With fallback on, unsupported AI systems continue to operate using fallback providers. Administrators can preview impact changes before saving configurations. Selecting entries in the Impact Summary links to the support matrix, which details AI systems, their activation status, types, and associated model providers, helping manage AI assets effectively. ## Audit logs Configuration changes within AI Control Tower related to Data, Approvals, and AI model providers are tracked in audit logs. Logs include timestamps, users, categories, and before/after settings and can be filtered by date (up to the last 90 days). In multi-instance setups, adding or removing managed instances generates paired audit records for synchronization tracking. AI model providers enable you to select data routing and manage third-party LLMs (Large language models) and SLMs (Small language models). ## Overview of AI model providers {#ai-model-providers__section_npn_npy_fkc} The AI Control Tower provides an AI model providers section where administrators can configure how AI requests are routed and which model providers are permitted on the instance. There are two categories of AI model providers: * AI model providers supported by ServiceNow. For example, Now LLM Service, AWS Claude, Now LLM-LTS (Long Term Stable) model. * AI model providers configured by your organization. For example, Perplexity, IBM Watson. {#ai-model-providers__ul_trd_spy_fkc} Note: You can select the Now LLM-LTS (Long Term Stable) model, which supports regulated industries such as financial institutions with stronger AI lifecycle management, governance, transparency, and conformance tools. For more information on the Now LLM-LTS (Long Term Stable) model, see [Long term stable models](https://www.servicenow.com/docs/elz6HbLpzCx8DuBQDNflkg "Long term stable (LTS) models support regulated industries, such as financial institutions, with stronger AI lifecycle management, governance, transparency, and compliance tools.") ## Data routing and model providers {#ai-model-providers__section_xyf_wpy_fkc} Data routing is a technology by which LLM and SLM requests are routed to the most suitable datacenter. This technology helps to optimize data traffic, which reduces latency and speeds up response time. The Data routing and model providers section enables you to route AI model requests and select the allowed model providers. There are two types of data routing: {#ai-model-providers__table_gcd_qzy_fkc__entry__2} | Data routing type | Description | |-|-| | Regional data routing | Routes LLM and SLM requests to datacenters within your region. For example, if you're in the APJC (Asia Pacific, Japan, and China) region, requests are routed to the most suitable datacenter in the APJC region. Regional data routing can be mandated by governments of specific regions. | | Global data routing | Routes LLM and SLM requests to the most suitable datacenter globally, regardless of the region where the request originates. | [Table 1. Data routing type] {#ai-model-providers__table_gcd_qzy_fkc} You can configure third-party LLM providers using the edit option by choosing either Regional or Global data routing and selecting all the Allowed model providers. ## Model Preview Program {#ai-model-providers__section_e2s_1sn_gkc} Model Preview Program (MPP) is an opt-in program that gives eligible users early access to AI models that aren't yet generally available. These models may be labeled as preview, beta, pre-release, or experimental. With the MPP, you can: * Experiment with upcoming AI models before general availability. * Test new AI capabilities in their environment. * Build and validate custom skills and agents using the latest models on the platform. * Provide early feedback on new model offerings. {#ai-model-providers__ul_lpf_rs4_gkc} How MPP works in AI Control Tower: * An MPP toggle appears in AI Model Providers settings. * The toggle is off by default. * An administrator must accept the terms and conditions before enabling the toggle. * When enabled, eligible preview models are set to available for use. * Acceptance is audited and recorded. ## Fallback and spillover {#ai-model-providers__section_q15_gqy_fkc} {#ai-model-providers__table_dmp_3yy_fkc__entry__2} | Feature | Description | |-|-| | Fallback | If you have active AI systems that aren't supported by your enabled model providers, the fallback mechanism enables these systems to continue operating with their default providers. If you choose not to enable fallback, these AI systems must be deactivated. AI systems deployed on fallback providers conform with the list of approved providers. The Fallback is activated by default and can be modified. | | Spillover | Regional deployments of AI models can experience limited capacity, which could lead to request rate limiting and impact performance. Enabling spillover can help prevent these performance issues. In ServiceNow, only Azure OpenAI currently makes this switch. The Spillover feature is set to active when Azure OpenAI is selected. | [Table 2. Fallback and spillover] {#ai-model-providers__table_dmp_3yy_fkc} ## Impact summary {#ai-model-providers__section_r12_kqy_fkc} The Impact Summary is determined by the chosen allowed model providers and the status of the fallback (active or inactive). The Fallback significantly affects how the Impact summary data appears in the Impact summary table. You can use the edit option to select **Yes** or **No** for activating the Fallback. Before saving, you can select Preview impact to review and confirm all your selections. **Activate fallback: No** When Fallback is set to No, the Impact summary table shows: * Total AI systems --- all AI systems supported by the allowed model providers. * AI systems supported by allowed providers --- AI systems with skill sets supported by the providers. * AI systems require deactivation --- all active AI systems that lack provider support and must be deactivated because the fallback option is not enabled. * AI systems can't be activated --- all systems currently inactive and not supported by any provider. {#ai-model-providers__ul_y5w_tyy_fkc} **Activate fallback: Yes** When Fallback is set to Yes, the Impact Summary table shows: * Total AI systems --- all AI systems supported by the allowed model providers. * AI systems supported by allowed providers --- all AI systems with skill sets supported by the providers. * AI systems supported by fallback providers --- AI systems that are non-conforming because fallback providers aren't permitted providers. {#ai-model-providers__ul_tbn_yyy_fkc} Note: The entries in the Impact summary table change based on the fallback status. When you select an entry from the table, the support matrix page appears with those selected entries, allowing you to update your personalized list. ## Support matrix {#ai-model-providers__section_xhb_dzy_fkc} The support matrix presents all AI systems in a table format, along with their respective AI model providers. The support matrix table includes categories such as AI system, type, activation status, and the selected AI model provider. When you select an AI provider, it appears in the AI systems and model provider support table. ## Audit logs {#ai-model-providers__section_ynl_3zy_fkc} Audit logs show configuration changes made on Data, Approvals, and AI model providers categories in the AI Control Tower Workspace. Select the View audit logs option to view the Audit logs page. The Audit logs page displays all configuration change details organized in the following categories: * Timestamp * User * Changed category * Changed setting * After change * Before change {#ai-model-providers__ul_l5q_jzy_fkc} You can also filter the changes by selecting a date range, starting with the last 90 days. Note: In a Multi-instance setup, when a managed (sub-prod) instance is added to or removed from the synchronizing instances, the audit log displays two records. The first record shows all instances being removed and the second record shows the instance being added or removed.