Many have dubbed this 12 months “the 12 months of AI brokers,” as these AI methods that may perform duties for customers are particularly helpful for optimizing enterprise workflows. At ServiceNow’s annual Data 2025 convention, the corporate unveiled a brand new mannequin in partnership with Nvidia to advance AI brokers.
Apriel Nemotron 15B
On Tuesday, ServiceNow and Nvidia launched Apriel Nemotron 15B, a brand new, open-source reasoning language mannequin (LLM) constructed to ship decrease latency, decrease inference prices, and agentic AI. In response to the discharge, the mannequin was educated on Nvidia Nemo, the Nvidia Llama Nemotron Put up-Coaching Dataset, and ServiceNow’s domain-specific knowledge.
The largest takeaway of the mannequin is that it packages superior reasoning capabilities in a smaller dimension. This makes the mannequin cheaper and quicker to run on Nvidia GPU infrastructure as an Nvidia NIM microservice whereas nonetheless delivering the enterprise-grade intelligence firms are on the lookout for.
The corporate shares that Apriel Nemotron 15B exhibits promising outcomes for its mannequin class in benchmark testing, confirming that the mannequin may very well be a superb match for supporting agentic AI workflows.
Reasoning capabilities are particularly necessary when utilizing agentic AI as a result of, in these automated experiences, AI performs duties for the top consumer in numerous settings. Since it’s performing duties with out human course, it must do some processing or reasoning of its personal to find out the right way to proceed greatest.
Joint knowledge flywheel structure
Along with the mannequin, the 2 firms additionally unveiled a joint knowledge flywheel structure — a suggestions loop that collects knowledge from interactions to additional refine AI fashions. The structure integrates ServiceNow Workflow Information Cloth and choose Nvidia NeMo microservices, based on the discharge.
This joint structure permits firms to make use of enterprise workflow knowledge to additional refine their reasoning fashions whereas additionally having the required guardrails in place to guard prospects, guarantee the info is processed in a safe and well timed method, and provides them the management they need. Ideally, this might feed into the creation of extremely personalised, context-aware AI brokers, based on the corporate.
Get the morning’s high tales in your inbox every day with our Tech Right now e-newsletter.
