How agentic AI is transforming building operations

As artificial intelligence moves beyond generating insights to making decisions and taking action, its potential to transform facilities management and industrial operations is growing. For manufacturers, agentic AI could help reduce energy consumption, improve equipment performance and free up teams from repetitive administrative tasks.

The Manufacturer spoke with Jamie Cameron, Vice President and General Manager of OpenBlue at Johnson Controls, who explained how the company’s AI-powered digital platform is helping organisations automate building operations, the importance of cybersecurity and what manufacturers should consider before adopting autonomous AI.

What is OpenBlue, and how does it support building operations?

JC: OpenBlue is our digital platform for digitalising the built environment, and I have end-to-end responsibility for the business, including product engineering, commercial delivery and customer success.

The platform focuses on three core areas. The first is asset performance, ensuring organisations get the most value out of their building equipment. The second is occupant experience, from providing applications for booking spaces to helping facilities teams service equipment more effectively. The third is sustainability, particularly reducing energy consumption and improving building performance.

What is agentic AI and how does it differ from the generative AI tools manufacturers are already  familiar with?

The starting point is getting data into a platform. We’ve built an end-to-end system that connects building equipment to the cloud securely, allowing us to collect the data needed to understand what’s happening across an environment.

Once that data is available, it can be analysed using different forms of AI, including generative and agentic AI.

The difference with agentic AI is its ability to move beyond analysing information and start making decisions. Rather than simply identifying an issue and presenting it to a person, the system can trigger a response and carry out the necessary actions.

For example, if equipment is operating inefficiently, the platform can identify the problem, make an adjustment and generate a work order where necessary. In some cases, this can happen without human intervention.

What can the OpenBlue AI chatbot do in practice?

The chatbot makes building data easier for people to understand and act upon. Traditionally, facilities teams would need to interrogate dashboards to understand what was happening. Now, they can ask questions in natural language, while the system can also flag issues and automate responses.

An example could be air conditioning operating in an empty meeting room. By analysing occupancy and meeting room data, the platform can identify that a room is being cooled unnecessarily, switch off the air conditioning and notify the relevant people through the chatbot. It can then restart the system when the next meeting is due to begin.

But there are also more complex applications. An energy analyst could ask the chatbot which two floors across a building portfolio are performing worst and the platform could analyse the data and provide an answer without someone having to compare multiple dashboards manually.

Equipment maintenance is another example. If a chiller develops a fault, the platform can identify the issue, check whether the affected part is under warranty and initiate the process of arranging an engineer visit. It could also bring forward an upcoming service and ensure the engineer has the necessary parts when they arrive.

This helps reduce unnecessary site visits and call-out charges, while making the process more seamless for facilities teams.

What are the biggest business benefits of moving from AI providing information to AI taking action?

The benefits include improved service, longer equipment life expectancy and lower energy costs. However, the biggest opportunity is the time saved by employees who would otherwise have to complete numerous manual processes to reach the same outcome.

Automating these tasks allows teams to focus on more valuable work rather than repeatedly investigating issues, raising requests and coordinating maintenance.

When processes are handled manually, things can also be missed or overlooked. Automation helps ensure that issues are identified and addressed more consistently, improving the overall performance of building operations.

Does this mean AI will replace facilities and operations teams?

I see AI as an opportunity to improve people’s working lives rather than simply replace them. Facilities managers can spend a significant proportion of their time responding to maintenance requests and equipment issues, which are often the parts of the job they enjoy least.

Automating these processes gives them more time to focus on work that genuinely improves the environment.

Buildings are still significantly behind other industries in terms of digitalisation and around 40% of global carbon emissions come from buildings. There is considerable scope to improve performance through better use of data and AI.

Think about how often lights, heating or cooling are left running when nobody is using a space. In commercial buildings, this waste can be substantial because energy consumption is less visible to individual occupants.

If we can reduce energy consumption by just a few percentage points across buildings, the impact on emissions could be significant. For me, that’s at the heart of what we’re trying to achieve with OpenBlue.

How do you approach cybersecurity and ensure organisations remain in control when AI can take action?

Cybersecurity is fundamental to our approach. Around four years ago, Johnson Controls acquired Temper Networks, a Seattle-based cybersecurity company.

Its technology creates a zero-trust network using micro-segmentation to restrict access to equipment. This allows us to isolate operational technology (OT) equipment within a secure, dedicated environment and protect communications between that equipment and the cloud.

The aim is to ensure that customers’ OT environments remain secure, with networks and ports protected against unauthorised access.

As AI becomes more capable of taking action within physical environments, organisations need confidence that their systems and data are protected. Building security into the platform from the outset is an important part of enabling that transition.

What have you learned from deploying the platform in real-world environments?

One of the biggest challenges is that no two buildings are the same. Even if two buildings have identical chillers, control systems and lighting, the way that equipment has been commissioned and configured can differ.

Before AI can analyse the data effectively, we need to extract it and organise it into a usable structure. This preparation is essential to getting meaningful results.

Continuous learning is equally important. The AI agents need to be trained on data from the specific building they are operating in, while the team needs sufficient knowledge of the building itself to understand what’s happening and extract value from the information.

The technology is important, but so is understanding the environment in which it is being deployed.

Where do you see agentic AI having the biggest impact across critical environments such as manufacturing facilities over the next three to five years?

I see AI increasingly operating in the background, with facilities teams benefiting from its decisions without necessarily needing to interact with the platform directly.

In mission-critical environments, for example, an AI platform could manage building operations, coordinate servicing and make decisions autonomously. The experience could become similar to a driverless car, with the system handling the processes required to keep everything running.

There’s also significant potential in workplace experience. Someone visiting a different office could simply ask a chatbot to arrange access, book a meeting room and organise the necessary facilities. The system would coordinate everything in the background, removing the need for multiple calls and messages.

We already have customers using these capabilities today, although adoption remains relatively concentrated in particular areas. The next step is to extend that seamless experience across more workplaces and operations.

What advice would you give organisations interested in adopting AI but not yet ready to let it take action autonomously?

First, experiment with your building data and understand what it can do for you.

OpenBlue takes raw data from a building and structures it so that people can query it using their own large language models. This allows organisations to ask questions they might not previously have been able to answer, as well as build dashboards and generate analysis much more quickly.

I’d encourage organisations to explore these possibilities and understand where the value lies before deciding how far to take automation.

Second, think carefully about the partner you choose. You’re opening up access to sensitive building data, so it’s essential to work with an organisation that understands cybersecurity requirements and recognises its responsibility for protecting one of your most valuable assets: your data.

Approach it as a journey and a collaboration with a trusted partner, rather than simply adopting a technology.