✕
Artificial Intelligence

ServiceNow’s AI Strategy Is Changing: What Should Customers Actually Pay For?

By Thomas Morgan

Like most enterprise tech companies, ServiceNow has spent the last couple of years building AI deeper into its platform, but recently it has also become much more open about letting other companies provide the tools and intelligence underneath it. Customers can now use third-party models from companies like Anthropic, Google, and Microsoft inside ServiceNow, while Action Fabric also makes it possible for external AI agents to interact with ServiceNow workflows from the outside.

On one hand, the move is sensible – most large companies are unlikely to bet everything on a single AI provider, and ServiceNow clearly wants to fit into whatever AI setup its customers already have (similar to Salesforce’s SIforce model). But if customers don’t actually have to use ServiceNow’s own AI, why should they pay for it too?

Although this doesn’t mean ServiceNow-native AI capabilities are irrelevant, the real value may be moving away from the model itself and towards everything around it, like workflow context, governance, and execution – which is where things get much more complicated, but interesting.

Does It Matter Who Owns the Model?

For a while, the obvious assumption was that ServiceNow would want customers using ServiceNow’s own AI models wherever possible, but this is starting to look a lot less important.

Since July this year, ServiceNow has been moving its out-of-the-box AI skills and agents over to third-party models by default. Azure OpenAI, AWS Claude, and Google Gemini were already supported, but the important change is that these models are no longer just alternatives. ServiceNow chooses them based on which performs best for a given use case.

Its own Now LLM Service is still available, so this isn’t ServiceNow getting out of the model business altogether, but it does suggest that owning the model is becoming less important than making sure the right model is being used by people.

ServiceNow competitor Salesforce is heading in a pretty similar direction. Agentforce can already run on different model options, including OpenAI, Anthropic models hosted through AWS, and Google Gemini. Salesforce also lets customers bring models from their own Azure, AWS, Bedrock, OpenAI, or Google Vertex AI environments into its AI platform.

The important part is that Salesforce still keeps its own layer around all of it. Prompts and responses can pass through the Einstein Trust Layer, which handles things like grounding in CRM data, data masking, auditability, and zero-data-retention agreements with model providers.

I make the comparison because both companies are coming to the same conclusion. The underlying model doesn’t necessarily have to belong to them for the AI product to belong to them too.

For ServiceNow, that means the Now Assist experience can still be a ServiceNow product even if Claude or Gemini is doing some of the actual generation underneath it. The workflow, data, permissions, interface, and governance can all remain in the hands of ServiceNow.

The Agent Fabric then takes this a step further by allowing agents to be built outside its platform (like on Claude, Copilot, or otherwise) to access governed ServiceNow actions. So the AI doesn’t necessarily even need to live inside ServiceNow anymore.

ServiceNow is saying customers can bring their own AI, not just their own model, while ServiceNow handles the workflows and execution around it. If that’s the case, the model itself starts to matter less, and the real question becomes what customers are actually paying ServiceNow for.

Is ServiceNow’s AI Actually Worth Paying For?

This is arguably where ServiceNow’s argument when it comes to paying becomes stronger. If it is happy to use models from Anthropic, Google, Microsoft, and others, then customers are not really paying for the model itself, but paying for everything it puts around it.

For companies, that’s still worth quite a lot. ServiceNow already knows the people, records, workflows, approvals, and rules sitting inside the platform. One person in an interesting Reddit thread regarding ServiceNow’s AI costs argued that this built-in platform context is the main advantage of using Now Assist, because an external AI tool usually needs much more of that information fed into it manually.

READ MORE: Now Assist: How ServiceNow Is Reshaping Work With Its Built-In, AI-Powered Platform

There is also the hassle factor. Building something yourself in Bedrock or Copilot Studio may be cheaper to start with, but then your own team has to keep it running, update it, test it, and fix it when things change. One commenter warned that this can quickly become another customer system that needs constant support. 

But the problem is that not every AI use case is that complicated. The original poster in the thread questioned why they should pay ServiceNow’s AI pricing just to help with something like employee onboarding. Another commenter made a similar point about basic use cases such as summarizing cases, where the cost can look difficult to justify if that’s all you really want.

This is where a dividing line emerges. If you want AI deeply tied into ServiceNow, with access to all its data and workflows, the native options start to make more sense. If you only want a model to summarize or draft something, and you already have AI tools elsewhere, paying for extra ServiceNow becomes a much harder sell.

So the question is less “Is ServiceNow AI worth it?” and more “What are you actually getting from ServiceNow that you do not already have?”

Final Thoughts 

There isn’t one clear answer to whether ServiceNow’s native AI is worth paying for because every customer is likely starting from a different place.

Some will already have big investments in Microsoft, AWS, Google, ChatGPT, or Anthropic. Others may prefer to keep more of the AI work inside ServiceNow because it is easier to manage alongside the workflows they already rely on. 

What makes everything more complicated is that Salesforce’s AI strategy is fairly complex at the moment. So many products, and so many ways to connect them, alongside third-party options too. While the flexibility is useful, it also gives CIOs a lot more to think about.

And with AI costs already front of mind for many businesses, deciding what should stay native, what should come from elsewhere, and what is actually worth paying for may be the hardest part going forward.

READ MORE: Ready or Not, Every ServiceNow Customer Is Now an AI Customer

The Author

Thomas Morgan

Thomas is a Tech Reporter at NowBen.

Leave a Reply