Introduction
Despite the futuristic caricature, Artificial intelligence isn’t that new. Researchers developed the early models of today’s algorithms in the 1950’s. What’s changed in the recent years is that AI has exploded forth from university laboratories to countless commercial applications.
In real estate, there’s been an immense amount of hype, but very little explanation of what AI is or how it works. Fortunately, like many things, while the details may be unbelievably complex, the core concepts can be explained to anyone.
To understand AI and why it now has practical use cases, we must explore a relatively new capability of AI: deep learning.
Fundamentally, deep learning is an AI algorithm that uses massive amounts of data from a specific domain to make a decision that optimizes for a desired outcome. It does this by training itself to recognize deeply buried patterns and correlations – many of which are invisible or irrelevant to human observers – to make better decisions than a human could.
The trick is that this requires massive amounts of relevant data, a strong algorithm, a narrow domain, and a concrete goal. If you’re short any one of these, things fall apart. Too little data? The algorithm doesn’t have enough examples to uncover meaningful correlations. Too broad a goal? The algorithm lacks clear benchmarks to shoot for.
So, in the example of building operations, we now have massive amounts of equipment data in the narrow domain of commercial real estate. What concrete goals can we select to identify hidden patterns to ultimately increase NOI?
1) Wrench Hours
Wrench hours, or the time it takes for operators or third-party vendors to complete a task, make up the lion share of maintenance and repair costs in commercial real estate. The fewer wrench hours needed to get the job done, the lower headcount you need, or the more value-add activities the same headcount can work on.
Seemingly small inefficiencies can add up to significant waste. Things like making multiple trips to find the right tools, searching for information such as O&M manuals or maintenance logs, and data entry can end up taking the majority of operators’ time on any given task.
With massive amounts of work order and equipment performance data, as well as the clear goal of reducing wrench time, AI has already come up with a number of applicable insights.
For example, equipment sensor data reveals nearly imperceptible signs of degradation as runtime hours between maintenance rack up. AI can balance the maximum acceptable threshold of this degradation with the time and cost of sending a technician to determine an optimal schedule. This can be taken a step further by grouping equipment so that work can be clustered and time maximized.
In addition, the AI decision-making can be coupled with other software, so operators are told exactly which tools to bring in the first place. The early indications show these AI-driven insights can reduce wrench hours by 75%!
2) Mean Time Between Failure
Most owners and operators understand that proper maintenance reduces the likelihood of equipment failures. But the type and cadence of activities can significantly affect how much of a difference is made with preventative maintenance.
For example, a maintenance schedule might state that air handling units (AHUs) should be visually inspected for proper operation daily, tested for proper belt tension weekly, have belt dressing applied monthly, and motor greased quarterly.
Obviously, the real world doesn’t work like that. Emergencies take precedence over routine maintenance and schedules get altered. By comparing real-world scenarios across hundreds of properties, AI can be deployed to solve for increasing the mean time between failure (MTBF).
The goal here is to maintain low wrench time while maximizing the effectiveness of maintenance that is performed. AI has even identified patterns that indicate that maintenance is being performed too often, leading to the cost of wrench hours without having any effect on MTBF. Delaying equipment repairs translates directly to an extended useful life and can represent many thousands of dollars in net present value for the owner.
3) Mean Time to Repair
Increasing MTBF is a valuable pursuit, but even the best AI cannot prevent every equipment failure. Unplanned repairs are just a fact of life in building operations. This is the other side of equipment useful life, the longer it takes to repair damaged equipment, the sooner a replacement will be necessary.
AI can be leveraged to determine the operational structure that leads to the lowest mean time to repair (MTTR). Owners have many options for how to manage their buildings. Some portfolios are managed entirely by outsourced facilities management, some are managed by a combination of in-house staff and maintenance vendors, and some are managed entirely by in-house staff. Even within those structures, there are hundreds of potential vendors and in-house resources can be deployed in numerous ways.
Leveraging equipment sensor data that could detect when equipment went down and when it came back online, AI could definitively state which structure leads to the fastest MTTR, and which teams within that structure are doing the best. This directly leads to solidifying best practices around a data-driven model, improved tenant comfort, and lower maintenance costs.
4) Energy Efficiency
To satisfy investor demands, more and more commercial real estate portfolios are imposing corporate sustainability goals that go beyond regulatory requirements.
The problem is that energy efficiency in buildings is seen (sometimes accurately) as coming at the direct expense of tenant comfort. Given the choice to save $50,000 by lowering the set point on a chiller plant or risk making tenants too hot, operators and property managers will always choose the tenant. Economically, this makes sense, keeping the building occupied will bring in far more money than will be saved on utility bills.
Fortunately, this is a problem tailored-made for AI. By inputting massive amounts of equipment energy usage, indoor and outdoor temperature, and ticket data, AI can be very precise in determining the ideal set points and schedules for HVAC and lighting equipment. By removing the false choice between tenant comfort and efficiency, owners can be confident they are pocketing the increased NOI from energy efficiency without risking vacancies.
5) Equipment Purchases
When it does come time to replace equipment, the process to do so is largely relationship based; decisions are often made because an engineer has worked with a certain manufacturer’s rep for decades.
But these are major capital investments that should be heavily scrutinized based on the upfront cost and the lifetime cost with utilities and maintenance included. Ideally, the decision takes into account the expected maintenance resources on site.
Wouldn’t it be great to feed the AI with historical operational data such as the MTTR of the property? Perhaps it is known that a property won’t have sufficient maintenance resources and it would therefore make more sense to pay more upfront to get a more resilient machine. This analysis would be impossible for any human, but with enough data, AI would be able to give a decisive answer.
Conclusion
The amazing this about AI is that it gives us a chance to reevaluate deeply held beliefs, assumptions, and rules of thumb. One drawback is that the process through which decisions are made is not accessible to humans. We know the inputs and get the outputs, but we’ll be blind to the thought process. Nevertheless, those who take advantage of this technology will be able to outcompete human decision making every time.
That is, as long as there’s enough data. The more examples of a given phenomenon a network is exposed to, the more accurately it can pick out patterns and identify things in the real world.
Every technology company will have access to computing power and software engineering talent. Once those have met a certain threshold, the quantity of data becomes decisive in determining the overall power and accuracy of an algorithm.
This pattern-finding process is easier when the data is labeled with the desired outcome. Building operators will continue to be needed to do the work and provide real-world feedback. Ironically, while AI can do computations that the smartest mathematician never could yet cannot do tasks that a child would find easy.
Enertiv has captured the most asset performance data in the industry, schedule a demo today to see we’re using AI to increase NOI for leading owners and operators.
.webp)
.webp)
