Digital workflow platform vendor ServiceNow is moving to fitting generative AI into every aspect of the enterprise workflow.
On Sept. 20, ServiceNow expanded its Now Platform with the Now Assist family of generative AI assistants.
The new capabilities are available in the Now Platform Vancouver release and include Now Assist for IT Service Management (ITSM), Customer Service Management (CSM), HR Service Delivery (HRSD) and Creator.
ServiceNow also released a domain-specific ServiceNow large language model, Now LLM, for enterprise productivity and data privacy. The vendor partnered with Nvidia to create its domain-specific LLMs.
ServiceNow's strategy of incorporating generative AI across all workflows differs from other vendors focusing on one or two areas, Futurum Group analyst Keith Kirkpatrick said.
"If you look at all of the different areas they're deploying AI technology, it really comes across everything from customer service to creating an application. All of these different aspects of their offering, they're incorporating generative AI," Kirkpatrick said.
For example, Now Assist for ITSM helps IT professionals with summaries of incident history and interaction with virtual agents that deliver complete answers to problems and requests.
Now Assist for CSM generates summaries for cases and chats, enabling customer service agents to resolve issues faster, according to ServiceNow.
Now Assist for HRSD summarizes case topics and context for HR professionals.
Now Assist for Creators includes text-to-code features and converts natural language into high-quality code suggestions.
Lara GredenAnalyst, IDC
The strategy of incorporating generative AI into varied applications also improves productivity for users because it reduces the probability of switching from one context to another, according to IDC analyst Lara Greden.
"That is the potential for imbuing GenAI into any workflow," Greden said. "Recognizing this, ServiceNow has long been making investments in unifying user experience and developing the AI behind ServiceNow Now Assist."
While generative AI-enabled workflows are not yet basic features of enterprise software platforms, Greden said she expects that will soon change.
"The early movers will be in the best positions to innovate with customers and create entirely new areas of value," she said.
Other than embedding generative AI into different workflows and use cases, ServiceNow's domain-specific language models show how the vendor is trying to make the best of not only its expertise, but also the expertise of others, according to Kirkpatrick.
As part of its generative AI strategy, ServiceNow provides customers with general-purpose LLMs, including access to the Microsoft Azure OpenAI Service LLM and the OpenAI API. Its new domain-specific LLMs are designed for ServiceNow workflows and are specifically for ServiceNow users.
For example, Now Assist for Search is powered by a ServiceNow LLM based on the Nvidia NeMo framework.
"It makes sense to leverage that expertise and the internal models on their platform," Kirkpatrick said. "And, of course, where appropriate, it's also good to incorporate other models for other use cases."
While other vendors might have similar strategies of using their own models as well as LLMs from different vendors, ServiceNow is specific and deliberate in saying that it will only use its models for specific processes to accomplish workloads or tasks, Kirkpatrick added.
The challenge with domain-specific LLMs will be whether they are successful, Greden said.
"It's early stages now, and thus, the low-hanging fruits are the areas chosen for these domain-specific LLMs," she said. "The challenge will arise with the next wave of use cases. We will see a proliferation in the number of LLMs and the associated need to manage them consistently."
Another challenge could be pricing, Kirkpatrick said.
Many companies, including ServiceNow, charge additional fees for enterprises to use their generative AI services.
"The challenge is going to be demonstrating real value," Kirkpatrick said. He added that specific tasks, such as summarization, might require less usage than others. "The question is, will enterprises find enough value there -- specifically when it comes down to ... one particular use case versus a more complex one."
Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.