In our previous post, we explored why EHR integration is foundational to deploying AI in surgical settings and walked through the process of how it unfolds. But connecting to the EHR is just one piece of the larger AI deployment puzzle. For AI to deliver meaningful insight in the operating room — whether it's tracking workflows, flagging inefficiencies, or supporting team performance — the underlying ambient technology also needs the right physical setup, and the machine learning algorithms need time to be trained from real-world activity.
In this post, we explore what happens after integration begins, how hospitals prepare their environment for AI, and how the technology starts learning about what’s happening in the operating room.
Who’s involved
Bringing AI into the OR isn’t a one-team job; it’s a cross-functional effort that spans perioperative, technical, and operational teams, as well as management stakeholders. Successful implementation hinges on early alignment and clear responsibilities across the following key players:
- IT Project Managers are the backbone of successful implementations. They keep the process moving, coordinate across teams, and ensure that schedules, materials, and communication stay on track.
- Facilities and Installation Teams handle the physical work: pulling cables, mounting cameras, and coordinating room access. They work closely with vendors and often install equipment during off-hours to minimize disruptions to surgical operations.
- Network and Security Teams ensure everything connects safely and securely once devices are installed. This includes firewall rule approvals, VLAN setup, and IP whitelisting. Early engagement with these teams is key to avoiding delays.
- Perioperative Leadership, including OR directors and nurse managers, help coordinate installation around ongoing cases.
- Customer Success and Field Engineering are present throughout the process on the vendor side, from walkthroughs and blueprint planning to model training and go-live support. They help install correctly, fine-tune the system, and guide hospitals through the onboarding and adoption phases.
Infrastructure setup
While AI is often associated with software and data, its rollout in clinical environments frequently starts with something far more tangible: physical installation. This process typically involves installing devices such as ceiling-mounted sensors or cameras across selected operating rooms.
Check out our case study on How Houston Methodist Increased Case Volume While Decreasing Overtime and drove a 10% case volume increase while cutting overtime by an average 36 minutes per OR per month.
Think of it less as "plug-and-play" and more as a carefully architected operation. Installation requires close coordination across facilities, networking, and perioperative leadership teams. The pace is dictated by room availability and hospital operations, with many institutions choosing to install in just two or three rooms per night during off-hours to minimize disruptions to care.
Before the first cable is pulled, the implementation team finalizes a detailed plan. This includes blueprint markups for equipment placement, procurement checklists, and a shared schedule. These planning documents outline every detail, including roles, timelines, network specifications, and hardware requirements, to prevent last-minute surprises that can delay implementation by weeks.
Done right, physical installation runs in parallel with EHR integration, setting the foundation for model training and eventual go-live.
Connecting the system
After the equipment is physically installed, the hospital’s IT and security teams step in to bring it online. This part of the process ensures that the new technology can connect securely to the network, communicate with other systems, and protect sensitive data.
While critical, this stage is also one of the most common sources of delay. Firewall approvals, network provisioning, and security signoffs can stretch across weeks, especially if key stakeholders aren’t looped in from the beginning. Misunderstandings about technical requirements or last-minute surprises can unexpectedly stall progress. Early and frequent alignment with IT leaders can prevent bottlenecks later on.
Teaching the system
Once everything is up and running, AI doesn’t immediately start delivering insights. First, the underlying machine learning algorithms have to be taught.
This training period, which typically lasts 30 to 45 days, involves the system observing how clinical teams operate, understanding typical patterns in case flow, and adapting to each hospital’s specific rhythms. In some cases, this involves AI models being reviewed and fine-tuned in the background to ensure they accurately interpret data and provide meaningful output.
It’s common for project teams to monitor progress during this phase and make minor adjustments as needed.
Avoiding the most common pitfalls
Despite the best planning, physical implementation projects for AI can still face delays. Some of the most frequent causes include:
- Limited Internal Bandwidth: Many hospital teams can only allocate a few hours per week to new initiatives.
- Infrastructure Surprises: Delays in ordering or setting up physical components can create unexpected blockers.
- Security Reviews: Multiple stakeholders may need to weigh in before systems go live, and late-stage feedback can reset timelines.
- Organizational Friction: New stakeholders, shifting priorities, or unclear communication can all slow progress.
Strong project leadership, clear expectations, and early alignment across all stakeholders are essential to ensuring a smooth and successful process.
Looking ahead
Once installation and training are complete, the real work of adoption begins. But who’s responsible for making that happen? While algorithms and infrastructure power AI, it’s brought to life by people.
In our next post, we’ll explore the human side of AI implementation: who needs to be involved, what roles are essential at each stage, and how strong cross-functional collaboration turns technology into meaningful results. Getting the right people involved early, across IT, perioperative leadership, facilities, and innovation, is often what keeps an AI project on track and prevents months of delays.
Check out our case study on How Houston Methodist Increased Case Volume While Decreasing Overtime and drove a 10% case volume increase while cutting overtime by an average 36 minutes per OR per month.