End-user Insight: Intel’s view on the digital transformation

Ryan Kovalchick is an engineering and automation manager at Intel. At the Ultrafacility 2025 conference, he moderated a digital transformation workshop focused on what makes digital solutions adoptable, where key gaps remain, and how facilities teams should evaluate technology providers. Kovalchick shared additional exclusive insights with the Ultrafacility team, offering an end-user perspective on what the industry needs to deliver going forward.

To what extent would you say digital systems and AI tools are being used in facility operations right now?
The semiconductor industry uses AI tools far more than other sectors like food and beverage. When it comes to facilities, however, digital transformation is still behind the fab. Operational technologies such as IoT and robotics are being used in the field to collect more data, not just from control systems, but from other sources as well, so that AI can then be leveraged to make more context-rich decisions based on that data.

So, I’m seeing progress, but adoption is slower because facilities are a more traditional environment compared to the high-tech fab space. Another challenge is encouraging people who have been doing the same things for decades to start considering new types of technology and integrating them into their day-to-day work.

Are you aware of any key providers who currently have AI systems available commercially for semiconductor?

There are providers offering advanced analytics solutions, but these are capabilities that can also be developed internally rather than purchased from a specific vendor. Over the past year, there has been a strong push from vendors traditionally used in the facilities space to start offering solutions with AI built in, even if it is including a chatbot to better analyse data and provide users with more accessible insights.

If you look at anything now, AI comes with it, and there are some imposters, but there are also real prime time solutions out there that are leveraging proper AI applications.

What would you say are the biggest operational challenges that AI has the potential to have a real impact in solving either now or in the future?

I think some of the biggest challenges AI can help with, both now and in the future, fall into two main areas.

The first is leveraging AI to control systems in a way that automatically rotates equipment on and off depending on the health of the system. Using AI to make those determinations is going to be a very big one. This is more of a command-and-control type application, where AI can take data from different data points, trends, and historical information and feed decisions back into the system. That could mean automatically rotating equipment or sending data from a controller to dispatch a robot into the field to inspect or manually manipulate equipment.

The other major area is maintenance. This is about transforming time-based maintenance into a more prescriptive approach, where you’re able to adjust work orders to avoid issues with your process. It also means you’re not wasting parts by replacing them too early and instead running equipment closer to the edge of where it’s meant to operate. That can save operational costs, as well as reduce time spent on maintenance and downtime.

What do you think the timeline will be for AI to reach the point where it is trusted to make decisions and actively implement them in facilities?
I believe we’re already at that point, but it really depends on the quality of the data being provided. From an intelligence perspective in the robotics space, I think we are within five years of having AI-enabled robots that are capable enough to handle a lion’s share of maintenance work, and even operational tasks.

Would you say that facilities are more interested in retrofitting AI into existing digital systems or investing entirely new off the shelf AI solutions?

It really depends on the business justification and the return on investment. Depending on the systems a facility already has in place, it may be easier to implement AI into legacy systems, particularly if those systems are tightly integrated into the business. That said, some companies may choose to invest in a separate, standalone AI solution. It ultimately depends on both the specific needs of the facility and the systems they already have.

Do you think the value of off-the-shelf solutions is high enough for facilities to invest in it and replace some of their existing infrastructure?

A key issue I’m seeing with AI adoption is the difficulty in proving business value. That’s partly because it’s hard to invest in capabilities when you don’t yet fully understand what they can do. As a result, many companies are struggling to demonstrate the value of the money they’ve already invested in AI applications, either because the technology isn’t delivering what they expected, or because it isn’t being leveraged in a way that creates real business value.

Ultimately, it depends on the capabilities of the solution and what it can actually do. If an AI system can replace functionality already provided by an existing system, there may be an opportunity to reduce technical debt - you don’t want multiple systems doing the same thing. At the same time, removing an existing application can be costly. This is all part of the broader architectural design of how organisations choose to leverage these types of solutions, and having a clear vision for that architecture is critical before moving forward.

How can the industry move from pilots and early-stage projects to commercial applications, and how can providers build trust in that process?

The most important part of a pilot is to vet it thoroughly and clearly demonstrate both the business value and the longer-term potential. Data is king, not just for AI, but for proving the value of any investment. A well-designed pilot should touch enough areas of the business to show meaningful impact, which is what enables organisations to move from pilot to deployment or commercial rollout.

If a solution is only piloted for a very narrow use case, it may perform well in that specific area but fail to demonstrate broader value. Having an appropriate sample size and scope is therefore critical to showing holistic benefits across different aspects of the operation, whether that’s safety, sustainability, operations, or maintenance.

To what extent do you think data privacy and cybersecurity are barriers to the semiconductor industry adopting AI-driven digital systems?

They are definitely barriers. From a data privacy perspective, organisations need to ensure that their data, whether it’s video feeds from technical solutions, robotics, or IoT systems, is properly protected, with the right safeguards in place to prevent it from being compromised.

From a cybersecurity standpoint, as more internal systems are connected and AI is layered on top, security becomes even more critical. If one system is compromised, there is a risk that vulnerabilities could propagate across interconnected systems.