Artificial intelligence (AI) is growing and evolving at a rapid rate. New applications can be found everywhere, including in the facilities industry. Evolving trends in data acquisition, data communication and storage, data science and artificial intelligence leave little doubt that smarter buildings are the future of the built environment.
These sustainable facilities will soon be self-sufficient, -actualizing, -maintaining and -repairing. Now is the time to prepare assets to reap the benefits of AI – benefits that translate into competitive advantages and result in a growing differentiation over time from competitors. This differentiation will help determine the relevance or obsolescence of tomorrow’s facilities management company.
One way to think about AI is in terms of operators and enablers:
AI operators are all the exciting, attractive executors of the instructions given by the artificial intelligence programs, including:
- Autonomous guided vehicles
- Smart applications
- 3D printers
AI enablers are everything else—the data, tools and processes behind the executors. They’re already being implemented in capital projects as preparation for AI operators and include an expanding variety of sensors, asset components and equipment. They capture data, real-time communication facilities, communication protocols for interoperability and, crucially, a holistic data model, channels and processes behind the scenes.
How to Prepare for Artificial Intelligence
The first step to embracing AI is to conduct a thorough assessment of your organization’s digital capabilities. This includes a holistic look at your people, processes and tools, and establishing a solid understanding of the current state of digitization. This will allow you to set intelligent targets. You’ll need to understand which data, assets, processes and paper-based files could be digitized.
A lot goes into the preparation stages. When you consider your data, here are three steps to take first:
1. Establish a data science function that considers the lifecycle of your data.
An initial and continuing emphasis is on data quality and character. Ask yourself how much you believe in the your data and where any gaps might be. A data governance framework that values and manages data with a holistic view of the enterprise can be the difference between long-term success and failure.
2. Address any concerns about data quality up front.
If it’s not, lower quality data will limit the effectiveness of machine learning.
3. Determine how to minimize the amount of data entry that’s left to humans.
Passive, non-invasive data collection is the goal. AI applications that involve interpreting large, complex data sets (like CCTV, audio and other sensor data) have matured and continue to evolve. Human interpretation of these types of big data sets can be time-consuming, difficult (depending on source) and prone to error.
It’s this type of unqualified data that AI can learn to interpret. Algorithms can be applied, for example, that allow a maintenance bot to scan for potential safety hazards, or identify and resolve discrepancies between BIM models and their physical instantiation.
Integrating an enterprise’s data into a Holistic Data Environment (HDE) is the challenge that lies immediately ahead.
Get ready for the future by developing your organization’s capabilities to manage and continually improve data transfer, storage and architecture. Every data silo is a data source. Identify your data sources and integrate them into a data warehouse that will serve as a repository for your data scientists and basis for machine learning.
Get the Most out of Artificial Intelligence Applications
Once you understand the quality, character and basic structure of your data, the next step is to think about potential business questions, and how your data may be relevant and provide insight into those questions. You will need to consider the suitability of your data to support various machine learning algorithms.
Different techniques have different data requirements. For example, a neural network generally needs a large set of data to train whereas other techniques (e.g., regression) potentially may be applicable to a smaller data set. Concurrently, you’ll need to consider the suitability of the data and techniques to the business question itself.
You may want to start with small data—data generated by enterprise resource planning (ERP) and management systems.
Where performance metrics are defined, target machine learning to understand relationships between ERP data structures and key performance indicators to uncover evidence-based insights. Knowing what good performance looks like unlocks advanced classification algorithms you can apply to your dataset.
There’s a confluence of creativity and analysis in conducting experiments and considering approaches to developing the AI pilot concepts—and there is a degree of uncertainty as well. Manage this uncertainty by establishing time constraints and go/no-go criteria for AI development projects.
The timeframe to put several AI pilots in place can vary greatly depending on your ambition and resources.
Take the time to develop a baseline understanding of your data. When it comes to smart buildings of the future, the possibilities for AI applications are endless. The right time to prepare for these possibilities is now.
About the Author:
Michael Goggin is an experienced software engineer and project controls expert. He applies systems engineering principles, cost control and project management principles to design and configure large, complex software implementations.