How and why IoT is shifting enterprise AI strategy

Sophisticated robots, drones, and automobiles on display at places like the world-famous Consumer Electronics Show (CES) are consumer playthings that, thus far, have won most of the media attention surrounding the IoT movement.

However, it is their industrial counterparts that are being used in tandem with artificial intelligence (AI) and machine learning (ML) techniques to transform operations in manufacturing, logistics, and asset management that will eventually have the most significant impact.

This is a hot market, with 8.4 billion connected things in use worldwide in 2017, up 31% from 2016, and with IoT spending forecast to reach US$1.29trillion worldwide by 2020.

Industrial Internet of Things (IIOT)

It’s not surprising that the industrial IoT (IIoT) revolution has commanded the lead in deploying IoT innovation and yielding returns from that investment, even if consumer IoT markets garner more media attention.

Industries like transportation, manufacturing, and utilities have long used machine-to-machine (M2M) communications (or telematics) for monitoring and control processes. The IoT extends the traditional operational benefits of telematics, enabling companies to use wireless technology and the Internet to connect a far broader spectrum of “things” to drive process improvements and product innovation.

The proliferation of powerful-yet-affordable sensors and data processors greatly expands the range of monitoring, control, analysis, and interaction of tasks that can be performed via thing-to-thing and thing-to-human communication. All of this means more gains (particularly when paired with enterprise AI) for cost savings due to operational efficiencies. 

The key to IoT & AI success – & what it takes to be ready

To succeed and remain a step ahead of traditional competitors, as well as IoT pure-players seeking entry into the market, enterprises will need to cultivate data as a core competency, using AI analytics. A survey from MIT of 1,480 business executives, managers, and IT professionals found that companies with robust analytics capabilities are 3X more likely to get value from the IoT than are those with weaker analytics capabilities. 

Developing data as a core competency means:

  1. Organisational change from the ground up, leveraging a combination of technology, people, and processes.
  2. Developing expertise in gathering, cleansing, integrating, and analyzing huge amounts of data. This must take place not only with data scientists but by leveraging and upleveling the skills of data analysts and those in the business sector as well.
  3. Understanding the importance of not just IoT data alone, but the power of combining sensor data with other sources like enterprise data (e.g., sales, CRM), third-party data (e.g., behavioural, demographic), and open public data (e.g., Census, weather), etc.

What makes IoT data unique

The types of information sensors in the IoT domain is highly diverse and expanding daily. “Sensors” refers broadly to devices that can detect, capture, and transmit (or store) information about their environment. This information, which may be transmitted in (near) real-time or recuperated periodically, increasingly bears a resemblance to the range of data the human senses can detect and process. 

As with humans, sensor data only has meaning once it is processed. Since there are billions of data points, data science, ML, and AI focus on an advanced interpretation of the aggregate data. This can help companies derive value from IoT data and make operations more efficient.  

Data science & machine learning in IoT is changing data science teams

Transforming extreme data sets into actionable insights and business innovation requires the right people, processes, and tools. IoT is changing the composition of data science teams. With the abundance of data from IoT, full data science teams working collaboratively are required instead of a single, jack-of-all-trades professional.  

With machine learning, algorithms can reveal patterns (descriptive analytics) or build models to forecast the future of behaviour of people, things, or processes (predictive analytics). Machine learning can be supervised (developed from a well-annotated sample set of data), or unsupervised (in which the computer mines data to discover patterns – or features – upon which to build a model on its own).

Overall, machine learning helps data scientists to filter the data, understand what is happening, and make predictions.

Traditional companies must adopt IoT to compete with each other and also with disruptive IoT startups. To find success, businesses will have to grow their data science capabilities in order to deal with all of this additional data both in terms of volume and complexity, developing more efficient ways to do ETL, transformation, processing, and modelling. 

At a technical level, successfully executing on IoT data requires adopting the platform, analytic techniques, and methodologies well-suited to IoT itself. By adopting IoT, companies will be able to bring significant value to their stakeholders and remain relevant.