Category: Machine Learning

The Commercial Real Estate Industry Needs to do More in Leveraging Machine Learning

pg_dmodml_052119

Machine learning and other cognitive computing technologies remain the hot, disruptive solutions marketed and touted by every software company out there, as the amount of money coming into AI-related startups continues to outpace other segments. However, the resulting “hype” has created a lot of unfortunate noise about machine learning’s value, and some commercial real estate leaders are having trouble navigating the buzzwords as they try to understand what this really means for them. Despite the noise, the power that machine learning can bring to an organization is absolutely real, clear and demonstrable, augmenting how we work and personalizing our experiences as consumers and employees. Instead of using rules-based programs, machine learning does just what the name implies: it learns from data and history to provide insights and patterns that are not found with the normal business intelligence and other standard, legacy analytic programs. On top of that, some of the most impactful results are those that leverage data that is sourced and specific to a company, building, employee or region, thereby providing personalized results that are unique and actionable.

A report by Deloitte last year revealed that early innovators leveraging machine learning, Natural Language Processing (NLP), computer vision and other cognitive technologies are already seeing business benefits. In the survey of 1,100 IT leaders, 55% said that their company’s adoption of AI has enabled them to increase their lead over competitors, with 9% stating that it actually enabled them to leapfrog ahead. That’s pretty compelling. So what about the commercial real estate industry? Is the industry taking advantage of the power that machine learning can bring? The answer is not yet.

This isn’t to say no one is leveraging any of the cognitive computing technologies—but its adoption has been limited at best. There are some wonderful machine learning solutions available today across all segments of the industry and I’ll name a few later, but first, why is the industry slow to adopt machine learning compared to other industries? Here are four big reasons:

  1. Data and knowledge intensive – The commercial real estate industry is fairly old compared to others, and it has historically been led by a relationship-driven, “run-on-my-gut” type of management, albeit with a relatively high success rate. That success rate feeds a legacy set of leaders that are skeptical of and resistant to change. Coupled with this, the industry truly runs on data but doesn’t have a great track record for compiling, housing, cleaning and leveraging that data. The legacy software applications were historically not open and did not support data integrations well, so consolidating the data took a lot of effort. This has been a major inhibitor in the industry moving as quickly as others in leveraging data and machine learning. That’s changing rapidly on the vendor side as new entrants provide more modern alternatives while pushing the legacy vendors, but there’s a lot of data already locked up in a myriad of systems across each company.
  2. Fragmented, with a myriad of legacy and modern applications – The industry has not attracted new software innovation in the past due to its heavy fragmentation (there are tens of thousands of owners, and even the biggest are relatively small compared to other industries). Though owning or investing in real estate sounds simple, the information needed to oversee and manage commercial real estate is fairly unique and broad (rent rolls, investor administration, complex lease terms, multiple regulatory agencies, tenant support, building operations, etc.) Therefore, it takes a collection of unique applications to meet end-to-end business needs, as there is no one application that truly meets all needs for every segment (commercial/multi-family, hotels, industrial, single family, geographical regions, etc.) On the positive side for machine learning, this fragmentation also adds to the volume of data that can be captured across the investment and operational lifecycle.
  3. Lack of attention to data quality – With all this data coming at every commercial real estate organization, very few have the data governance maturity needed to yield high quality data. A recent poll that I saw of fellow commercial real estate CIO’s showed that one of the biggest hesitations in moving forward with machine learning and similar technologies was a lack of confidence in their data quality. It’s not the only reason, but most CIO’s rightfully realize they need to have good data in order to achieve the best results.
  4. Resource constrained – On the earlier point that the industry is primarily made up of small- to medium-sized companies- this factor translates to organizations not having large employee bases, so they are inherently resource constrained. However, that is the catch-22 about machine learning, as it can make employees more efficient by automating the more mundane, operational tasks, freeing up more time to focus on customers and knowledge-based tasks. The employees will also be armed with new, data driven insights to carry out this work.

Ok, so now that we’ve reviewed why the industry has been slow to adopt, let’s focus on nine ways that machine learning can add business value to the commercial real estate industry, and some examples of companies that are already bringing value:

  1. Proactive and predictive insights on asset conditions and failures: One valuable use of machine learning that is getting traction and validation is the ability to more efficiently operate and manage a building’s physical assets and IoT devices. Specifically, the ability to convert the fault alerts produced by building equipment into actionable notifications, providing early warnings when important assets might be failing or need attention. There are a number of companies that filter through these warnings based on patterns and historical resolutions, while some are also looking more broadly to include other data points—such as work order history or manufacturer/device history—to identify and predict when an asset is in danger of failing. Fixing or replacing a piece of hardware proactively is much more cost-effective and incurs less employee impact, so predicting when an asset might fail can be beneficial. Some of the companies addressing these issues are BuildingIQ, Enertiv, and Switch Automation.
  2. Gaining proactive insights into tenant space needs, operational issues and other factors affecting NOI: Most of the “insights” available today are via operational applications or traditional business intelligence tools. That’s great for understanding what happened in the past, but these tools are not ideal for helping predict what might happen next. They are also not good at leveraging the disparate sets of data available in providing proactive property and tenant insights. By combining both internally sourced data (workorder, lease expiration’s and terms, parking, a/r, etc.) and external data (weather, tenant growth metrics, etc.), Okapi is one company actually using machine learning to provide insights in this manner for commercial and multi-family, while diffe.rent and home365 are a few examples in the single family space.
  3. Occupancy and space utilization, and the personalization of the workplace: Understanding an office’s space utilization patterns is one of the most impactful, but less optimized functions of a corporate real estate organization, with the biggest issue being the ability to easily understand how office spaces get used on a daily basis. With COVID, this topic has become even more important and critical to ensuring a safe Return To Office.  Some of the solutions being deployed are leveraging Computer Vision to ensure employees stay at safe distances and to confirm that occupancy levels are at a safe level. Understanding and predicting usage is also important to ensure that space is safe, while also supporting employee productivity and identifying future opportunities to support growth. Machine learning can analyze a disparate set of data coming from sensors, room booking, badging, and other siloed sources and highlight usage patterns that might be unique to a specific building, region or department. A few companies playing in this area are Digital Spaces, Density, and Vergesense.
  4. Enhanced tenant and employee engagement: One of the biggest trends in commercial real estate is the explosion of employee and tenant-facing apps that aim to connect users to the services and communities that matter most. They support conference room booking, facility and work order requests, class registrations, ride hailing, cafe menus, and many other features that are increasingly expected by today’s employees. The more sophisticated apps leverage machine learning and historical data to suggest specific conference rooms, advise of non-bookable working spaces that might become open, or recommend class registrations or specific parking spaces, among other tasks. It may not sound like much, but any friction or key strokes you can remove from an employee’s day go a long way in their job satisfaction. Some examples of these technologies are CBRE’s Host, Workwell, and HqO, to name a few.
  5. Insights into property valuations and buying opportunities: The selling price for commercial real estate has many factors, so determining the best value for an asset, or highlighting underpriced assets, are great examples of where machine learning can add value. Most buyers use discounted cash flow and other financial models to help determine an asset’s current value, so the more accurate an assumption is on rent growth, occupancy, and market rents and demand, the better the valuation model will be. In 2018, there was more than $562 billion worth of commercial real estate transactions in the U.S. alone, and this large transaction volume offers a treasure trove of data and information. It’s a lot easier said than done, but companies like skyline and others are developing machine learning algorithms that offer investors and partners access to the sophisticated insights machine learning can offer.
  6. Computer vision: Computer vision is leveraging machine learning against images and videos for insights, and it’s an area that is early but one that will be very transformative for real estate over time. Computer vision is also used by robots that can navigate and monitor both indoor and outdoor spaces in various ways. There are many use cases in production today (I’ve lumped them together for simplicity) such as occupancy counts, identifying demographics of shoppers, security notifications on crowd gatherings, license plate and visitor blacklisting, employee building access, employee or tenant sentiment, and even early warning notifications to law enforcement when an active shooter first brings out a gun. Though there are real concerns about privacy when not used properly (a larger topic on AI ethics that needs its own summary), the technology is there and already in use. Companies like trueface, aegis, Knightscope, ambient ai, and Cobalt Robotics are just a few examples.
  7. Automating the lease abstraction process: Real Estate is one of the most document-intensive industries, so it makes sense to leverage machine learning to automate some of these unique processes. The lease abstraction process in particular, is manual with the non-standardized use of leases across the industry, along with the variation of terms and clauses found in every lease. By utilizing NLP, a form of machine and deep learning that analyzes words and context from history to take actions, the lease abstraction process can be augmented to improve efficiencies and lower expenses. Leverton was one of the early pioneers in the industry and were recently bought by MRI, while DealSum is another. However, most of the cloud platform players have advanced NLP capabilities, with Google being one of the leaders in this arena with their Document Sense product. We are very early in this segment, as labeling is complex and time consuming, but it is one that has high ROI opportunities for the larger firms with a high volume of leases.
  8. Automating the work request process: Most facility management applications can be cumbersome and time consuming to create a ticket, since many of the applications require multiple inputs (location, request type, urgency, description, etc.) to process and assign the ticket. To improve and simplify the user experience, NLP models can leverage the words used in historical requests to automate the process, requiring only a basic description of the problem. Machine learning programs can learn from the language used to describe a problem, taking words like “water leak,” “broken handle,” “coffee spill” and other words used in previous requests to assist and automate the creation and assignment of the ticket. Most work order systems today don’t yet have this capability built in, but I’ve personally been involved in the development of similar work efforts that leverage some wonderful machine learning platforms like Google’s GCP AI products and Microsoft’s Azure AI.
  9. Leveraging chatbots to interact with tenants or employees: This last example of machine learning is actually one that is the most pervasive and real use case across all industries today. Known more formally as “conversational AI,” chatbots and virtual assistants leverage machine learning and historical data to automate the most redundant, typical and time-consuming requests carried out by employees. You’ve likely come across a chatbot while visiting a website, or maybe you’ve “chatted” with “someone” via web support, when in fact it very well could have been a chatbot. The beauty of a chatbot is that it’s always on and waiting, and it can handle the first level interactions that cover the majority of requests. Developed properly, they can escalate the issue to a live person if a question isn’t being answered or upon request by a user. In commercial real estate, chatbots have been deployed to answer tenant questions and resolve facility issues, with just two examples being the Bengie app from Building Engines and CBRE’s host.

This is just a short list of some of the machine learning use cases and companies being leveraged today, but I hope it provides insight into what is possible and the business value that machine learning can provide. Over time, these and other capabilities will become more mainstream in the commercial real estate technology world. For now, it’s the early innovators that are ahead of the game and leading the pack.  Are you one of them?

AI Impact on Reliability and Safety Within the Energy Sector

CBRE_Article-2

Article originally posted in Realcomm Advisory Newsletter

Ride sharing apps. Movie and shopping recommendations. Social media feeds. Virtual assistants. Music and media streaming services. You have likely used at least one form of artificial intelligence (AI) already today, as it has become ubiquitous in our everyday lives. By definition, AI is simply the science of training machines with data to mimic cognitive functions and perform human-like tasks.

The use of AI – and its ability to analyze and leverage large amounts of data – has risen to the forefront of corporate strategies. Examples of current uses include harnessing real-time building data to proactively predict asset failure; measuring temperature and motion to create better experiences; robotic use for material handling, delivery and construction; facial recognition and security; and performing lease abstraction by extracting relevant data from piles of hard-copy documents.

For some, the future of AI is exciting – for others it may be daunting. But, the “future” is now, and there are remarkable strides being made with everyday applications of AI that are helping us work smarter, faster, safer and more creatively by automating manual tasks and supporting our decision-making process.

Learning from Structured and Unstructured Data
AI is only as good as its data. To be truly effective, AI and machine learning require large amounts of data that is both diverse and clean. Across the real estate industry, there is a heightened focus—and an increasing amount of investment capital—around getting our data together, normalized and available for use. At the same time, the amount of data coming in is exploding, particularly unstructured data that is created via cameras (images and videos), audio and speech recognition devices, mobile content, web pages and documents.

Gartner, a leading global IT research and advisory firm, predicts that data volumes will grow 800% over the next five years, and up to 80% of that data will be completely unstructured. It’s also estimated that only 1% of this data is ever analyzed, so capturing, preparing and creating insights from this data is a challenge that many in the industry are unprepared to tackle without the help of AI.

Energy Sector Leveraging AI
The energy sector has recognized the value of AI across many facets, including energy center operations, reliability, and safety. When it comes to pipeline inspection and leak detection, AI is helping to prevent issues before they arise. For example, intelligent extraction robots are already in-place today, improving cost-effectiveness and productivity while minimizing worker risk. In addition, advanced Leak Detection Imaging Systems, which leverage neural network-based AI, are using algorithms to process images and detect, confirm or reject potential leaks.

By using data and identifying patterns, companies within the energy sector are also increasing their capabilities around cybersecurity, threat detection, and the optimization of assets and maintenance workflow. In developing countries, electricity theft can be prominent. Machine learning can help utility companies identify suspicious patterns and determine normal usage rates of residents. Any inconsistency can be identified, monitored and reported, to avoid further electricity theft.

As a service provider, improving asset maintenance and both the employee and the customer experiences can benefit from data and machine learning. On the employee experience front, CBRE recently completed a pilot that automated the CMMS workorder request process, using Natural Language Processing (NLP) to identify, process and extract the employee intent and request from just a basic text string. That same process will also be leveraged as a chatbot (aka conversational AI) in a future product release to our CBRE 360 clients.

Companies within the energy sector can also leverage chatbots to improve their B2B interactions and customer service capabilities with resellers and vendors. There are additional, AI-enabled pilots underway that will allow energy sector companies to better predict asset failures, and also optimize technician routes by leveraging workorder data in conjunction with weather and traffic information.

Future Vision
The AI space will continue to evolve and influence multiple aspects of the energy sector. According to Dan Walker, who leads emerging technology in British Petroleum’s (BP) Technology Group, “AI is enabling the fourth industrial revolution, and it has the potential to help deliver the next level of performance.” As new technology is released, and the use of AI is more widely adapted, you can expect to see considerable advancements that will revolutionize the way we work in the energy field.

The Future of CRETech is Here

I recently had the privilege of participating in CoreNet’s Technology Symposium, hosted by CBRE’s Peter Van Emburgh, President of the Mid-Atlantic chapter in Washington, DC, along with Katy Redmond. The theme of the symposium was the “Future of CRE Technology.” Being in the middle of some innovative technology projects at CBRE, this is a topic near and dear to my heart and I was thrilled to be involved. At the symposium, I participated on a panel with some great minds from IBM, Saltmine, IA Interior Architects and Capital One, and there was broad agreement among the panel that IOT, Machine Learning, Blockchain and Augmented/Virtual Reality continue to be the hot topics on innovation in Commercial & Corporate Real Estate Technology (CRETech). Below are some themes we discussed and my views of what this means for our industry.

CRETECH INVESTMENTS AND TRENDS

The real estate technology world, though still behind other industries, has come a long way in recent years. The amount of capital being invested in this space continues to be extraordinary, hitting $3 billion in 2017 across all real estate verticals. That’s on top of the $2.7 billion invested in 2016 and $2 billion in 2015. That’s a lot of money coming into real estate tech and we’re now seeing more and more startups bringing innovative, digitally focused, intuitive, and disruptive products to the industry. It’s still debatable on whether there is too much money being invested in an industry where exits are still rare, but it’s been a great push for moving CRETech into the middle of the digital revolution.

The keynote was given by Steve Weikal, the Head of Industry Relations at MIT Center for Real Estate, who gave a great overview of what he’s seeing from a trend and startup perspective. On some of the broader trends impacting CRETech, the pervasiveness of cloud computing, mobile computing, the explosion of “Big, small and wide data,” and the increasingly tech-savvy CRE workforce stood out and resonated with what I’m seeing. Along with the sharing and on-demand economy, these are all foundational trends that have had positive impacts on CRETech.

IOT

From a near-term practical perspective, the Internet of Things (IOT) is finally being broadly adopted and leveraged by real estate and workplace executives. IOT powers many enterprise and consumer products, and for the CRE world the Smart Building concept is not new. It’s been a topic of discussion and attention for over 15 years, but there was always more of a “when” attached to it than reality for most companies. That’s no longer.

The accelerated decrease in sensor prices, the ubiquitous view of mobility, the broad availability of cloud storage and services, and the ongoing adoption and capability of API’s, has resulted in IOT & Smart Buildings becoming ever more pervasive in CRE.

Energy management and related savings were the primary beneficiaries of the initial IOT waves, but we’re now in a world of “experiences” –one where IOT devices are star players. Adding Machine Learning to the mix, you now have the personalization of the experience. From conference room booking to wayfinding, people finding, parking and transportation search and reservations, service requests, community interactions, food or concierge services, IOT enables new ways for employees to interact with their workplaces.

 MACHINE LEARNING

Machine learning is another powerful advancement in the world of CRE tech. When Machine Learning is applied to IOT, the ability to identify and understand how space is utilized magnifies greatly. With a minimal number of sensors, coupled with machine learning, companies can now understand how employees and departments are actually using their space, adopting new working styles and improving the efficiency and effectiveness of their real estate portfolio. The combination of IOT and machine learning also can provide insights into asset performance, event filtering, and service work validation. These are all projects that we’re piloting with our clients with goals to improve performance and increase employee satisfaction.

The value of Machine Learning to the commercial real estate industry is broadly apparent. Its ability to find insights in data are transformational. The big tech companies (Microsoft, Amazon, Google, IBM) have all made huge strides in making machine learning models and approaches more available to a broader audience, “democratizing” the availability and use of machine learning. Related to the workplace experience theme mentioned earlier, we’ve spent time at CBRE applying machine learning to work-order data, enhancing and improving how employees can get what they need to do their job. Whether it’s leveraging text, voice or images, making the employee’s interactions simple, easy and personalized goes a long way towards employee satisfaction. That is something critical in today’s highly competitive fight for top talent.

BLOCKCHAIN

Usually blockchain in CRE is a topic for a narrow niche group at CRE technology meetups and conferences, but it got much more air time at the symposium than I’ve seen before, which is great. There is no arguing the value that a decentralized, secure, and fully searchable technology like a blockchain can have on the CRE world, but there aren’t many corporate or commercial real estate companies who are prepared to leverage it today. At CBRE, we completed a successful proof-of-concept with a client last year that opened our eyes further to the long-term potential. There are quite a few uses today that are viable in the short-term, particularly around logistics or IOT, but other aspects are many years away. Outside of the U.S., there are governments that are pushing to centralize all real estate transactions on a blockchain, enabled by a lack of available data today and a supportive and centralized regulatory approach. In the U.S., Cook County and other municipalities have piloted title transfers on a blockchain, but broad implementations are yet to come, particularly on the commercial side. Some of the best use cases require multiple organizations to participate, and many aren’t ready. Still, that shouldn’t stop organizations from exploring how it works and its benefits. I’m confident it will be a very disruptive technology over time.

AUGMENTED AND VIRTUAL REALITY

We also spent time on the panel talking about augmented, mixed and virtual reality, and it was agreed that this is a technology that is perfect for real estate and the workplace. Companies like Floored have already blazed the trails in creating innovative, virtual walkthroughs of space, but that’s just the beginning. Augmented reality has a real potential today that isn’t yet fully leveraged. Imagine a building engineer being able to see the work order, warranty or other documentation about a specific asset, just by holding their phone to in front of the asset. With AR being incorporated into the new phones, that’s starting to happen today, improving the time it takes for building engineers to do their work and reducing the time and cost of getting remote support for solving problems.

Since virtual reality is more about being fully immersed in your virtual surroundings, there is a thought that it would supplant the video call, allowing remote works to “feel” like they’re all together. That concept was discussed but there wasn’t a general agreement on its adoption or value. Completely removing the physical nature of human interactions is doubtful, but the current video-meeting experience is not the best either and ripe for disruptive changes.

CONCLUSION

In summary, we’re in a great time where the advances in CRETech are real and here now, with a bright future ahead for IOT, machine learning, blockchain and augmented reality. These technologies will all make a big impact in the employee experience, and it’s about time.

If You’re Not Leveraging or Considering AI in Some Fashion, You’re Already Behind.

It’s been said that data is the new gold. If you believe this critical concept like I do, then you should believe that artificial intelligence (AI), and machine learning in particular, will have a tremendous impact on the way we work. Though not new, machine learning is the next step in the evolution of data analytics and every company should already be looking at where they can take advantage of this transformational technology. If you aren’t doing that already, you’re behind.

Data Analytics too often is about looking backwards at what happened, while machine learning is about looking at the same historical data but with an eye into the future. It uses patterns in the data to help provide insight and recommendations and it can help automate tasks where historical patterns are a good prediction of the future. Machine learning is just one application of artificial intelligence, but it’s the most accessible and relevant as it applies to most enterprises. There is a lot of debate about how robots powered by AI might replace our jobs, but we’re still too early in the AI evolution to be worried about full scale replacement. Many firms are already taking advantage, augmenting their employees work and freeing up time for more knowledge based activities. It’s a long way before large sets of jobs are replaced, but AI can and should be used to augment and improve processes, while also enabling personalization into how each employee or consumer works or shops.

Machine learning also opens up the world of prescriptive analytics, which takes predictions a step further by suggesting actions based upon the predicted event. Just because you know something will happen doesn’t mean you’ll take the most meaningful action. Scale also becomes more attainable with AI. The augmented work, insights and predictions open a world of doing much more with less.

If you’re still figuring out how to get started, then don’t worry. Foundationally, it’s much easier today to start than it was just one year ago. The cloud has accelerated machine learning adoption and accessibility and it’s a perfect environment for machine learning. You can get a lot of compute for specific periods of time for model training, while the number and quality of the model’s available increases daily. All the major cloud vendors now make it easier to tap into their models no matter where your data resides, while also providing open API’s for integrating the results into usable forms. There are also a multitude of vendors available that can help you start with a small project as you begin your learning exercise. Like other new technologies, it’s smart to start small in the form of testing, learning, trying and discovering.

When looking at where to begin, think about what business problems could be solved by automating a routine task. Look at problems where understanding historical trends can improve decision making. As you approach the project, keep the Agile methodology in mind; identify the business problem, pick a short-term win that can be accomplished in 4-6 weeks. Take the Minimal Viable Product (MVP) approach and don’t try boiling the ocean on this. You need to try it, learn, iterate, and go through it again.

Some examples of where to look are:

·       Repetitive tasks – Are there tasks that staff perform that are routine that also generate transactions or other data sets? Do you get requests that are repetitive and where history is a good indicator of how to process these requests? Is there data movement between systems that are routine?

·       Providing insights – Are there key processes that are event based, where you might improve the outcome the next time if you had historical data that provided insights into the quality of your decisions?

·       Reducing Noise – Operationally, do you have notifications that generate more data than you’re able to easily sift through? Machine learning can help you get through the noise and clutter.

·       Improving the user experience – Are there processes in your operations where your employees are required to go through multiple steps to get help or generate requests? There are many use cases where machine learning can help employees get what they need quicker, easier, and cheaper.

·       Personalization – Leveraging user preferences and habits, the user and employee experience can be personalized to provide a better and engaged experience. What temperature do you like it in your workspace? Which conference room do you tend to use (or what is the most convenient). What food do you like the most, giving you alerts when it’s available in a nearby café?

·       What’s in your data? – Look at where you have a lot of data. Just looking at the systems or devices that are generating a lot of data can give you ideas for where to look. The more data you have to train and test the model is important, so start with what drives your current data analytic requirements and you’ll likely get ideas from there.

One last note about your data. I’ve written many times about data quality and the importance of data governance, and the topic of machine learning is a great example of why that’s critical. Hopefully you have a good data governance program in place and your data is somewhat clean. Data quality will be key to a successful AI pilot. Having said that, you also don’t need perfection to start. Just by starting an AI project will give you insights into your data and hopefully get you on a journey of organizing and improving the quality once you understand what you really have.

If you haven’t thought about starting an AI program, then you better start soon. I bet your competitors are already on the journey and you’ll soon be left behind.