ServiceNow, Anthropic Partner On Using Claude For AI-Powered Applications

‘We are bringing our domain and knowledge as well as context to build applications using Claude’s models trained on that industry-specific content to build a differentiated offering,’ says ServiceNow President, COO and Chief Product Officer Amit Zavery.

ServiceNow Wednesday unveiled a collaboration with Anthropic to tighten the relationship between Anthropic Claude AI models and ServiceNow workflows.

Under the agreement, Claude becomes the default model behind ServiceNow Build Agent, the company’s vibe coding technology that uses language reasoning models and retrieval augmented generation (RAG) to transform plain language and business needs into code.

ServiceNow President, COO and Chief Product Officer Amit Zavery told CRN that his company has been building out its agentic AI platform and delivering lot of automation to customers in partnership with multiple vendors and choice of cloud systems and large language models.

“We also have seen some unique use cases emerge where we might be able to do more with a particular partner,” Zavery said. “In expanding our relationship with Anthropic, we have a capability called Build Agent, which we released last year. It's a vibe coding interface. We’ve always had no-code and low-code to build and extend our ServiceNow platform. Build Agent provides a vibe coding interface to build a workflow or prompt.”

That code then can run on the ServiceNow platform with all the enterprise compliance, security management, runtime guaranteed performance, and so on, Zavery said.

“And for that, we are starting to more and more use Claude as the underlying code generation tool and build agent,” he said. “And we added lot of capabilities around enterprise security and things like that to make it work very fast but also guarantee full enterprise compliance capability.”

Zavery said customers are free to use other platforms other than Claude for the vibe coding.

“One thing we’ve seen with Claude is it does have a very good code generation capability and the ability to build and generate code based on prompts,” he said. “We've been using it inside ServiceNow for some of our engineering work. This is a vibe coding environment versus low-level coding, but we know the generation is pretty good and powerful, and we found we can make it even stronger with our Build Agent.”

When asked about issues businesses have faced with vibe coding such as security risks, code quality, scalability, and so on, Zavery said Claude with Build Agent mitigates those issues.

“When you build an application, we provide you the data model,” he said. “When building a workflow, we generate a lot of the data model based on our historical way of building workflows. So the grounding and scaffolding around these models really prevents a lot of those errors and hallucinations. That, I think, is a big difference when using Build Agent and Claude versus Claude by itself, or any other code generation tool out there for enterprise use cases. What we have done with the platform is track what the application looks like. How does it create it? What is the data model? We are really guaranteeing that outcome based on our domain understanding and expertise of many, many years.”

ServiceNow is also seeing a lot of interesting AI use cases emerge, in particular in regulated industries including life sciences and healthcare, Zavery said.

“Anthropic with Claude has done a lot of training of data sets with our domain and context understanding of specific industries,” he said. “We’re bringing those together to co-innovate on things such as issue resolution, research analysis, things you want to do for case management, claims, and things like that. We are bringing our domain and knowledge as well as context to build applications using Claude’s models trained on industry-specific content to build a differentiated offering.”

Furthermore, Zavery said, ServiceNow last year launched its ServiceNow AI Control Tower to manage the lifecycles of AI deployments, AI agents, and AI systems, Zavery said.

“We discover it, we manage it, we observe it, and we track it,” he said. “We help with the lifecycle as well as security and the data associated with that from the outcome perspective,” he said. “We are also now able to discover Claude, easily track it and make it part of the ecosystem, so customers can manage all their AI implementations and lifecycles through one product called AI Control Tower. It’s built on CMDB (Configuration Management Database) and extends very well. We do all tracking of hardware and software. We add AI for tracking and management, and that makes our product much more powerful to allow customers to make sure all the products and things they might be using are supported through our AI Control Tower.”