Stanford Health Care is implementing agentic AI to improve network management across its extensive system of over 200,000 devices.
By streamlining operations and reducing the manual workload for engineers, agentic AI seeks to enhance overall network performance and user satisfaction within the health care environment.
“Agentic AI isn’t just answering questions, it’s like adding a digital teammate who can take on complex, multi-step work. In health care, this is especially valuable because any network delay can affect patient care,” explained Ahmad Rezazadeh, Manager for Network Engineering.
Unlike traditional methods that respond to issues after they occur, this AI proactively monitors the network of connected devices across 60-plus SHC locations and 3.5 million square feet of floor space. It identifies problems and suggests solutions, helping to minimize disruptions that could impact patient care.
“It can start investigations on its own, connect the dots between systems, and even draft communications or fixes. That means less manual detective work, faster resolutions, and fewer disruptions for staff and patients,” said Rezazadeh.
The NetSmart AI agent now does much of the legwork automatically — gathering logs, identifying root causes, and even drafting updates for engineers, cutting down troubleshooting from hours to minutes, he explained.
Rezazadeh’s team is focusing on high-value, repetitive tasks where AI can make a big difference:
- Wired network incident resolution: Automating the way we find and fix cable/switch-related problems.
- Configuration compliance: Making sure devices follow security and performance rules automatically.
- Firewall and load balancer changes: Agents prepare, validate, and sometimes execute these changes, reducing errors and approval delays.
- WAN link monitoring: Watching our wide-area connections in real time so we can prevent outages instead of reacting after the fact.
“The biggest benefit is going to be in identifying the issue proactively before user reports it, and then solve it proactively or reach out to the user to let them know that we have identified the issue he is facing and actively working on to resolve it,” said Rezazadeh.
“We should be also able to do more with less as resources can be freed up to do more project-related tasks where the network team member needs to go onsite for network build activities which can’t be automated yet — I mean, not until robots take over!”
