IoT Vs. Embedded Intelligence: What's The Difference?

I recently read an article in the New York Times about the disappointment most people feel when they use sleep trackers. Basically, people have spent good money on high-tech sleep trackers, but the devices simply tell them what they already know—they aren’t sleeping well.

The customer thought they’d receive some sort of help getting a better night’s rest, but instead, they just got a data collection device and an app that monitored how long and how deeply they sleep each night.

This is a common problem in the world of IoT, or Internet of Things.

Customers are expecting solutions to their problems, but most IoT products are rather passive. An IoT device is nothing more than a device that has integrated sensors to collect information, and connectivity to communicate this information. Turns out that actually creating operational intelligence is not that easy.

Far too often, both the general public and companies who engage with smart technology conflate the term “IoT” with something else altogether: embedded intelligence.

Embedded intelligence and IoT are not the same technology. And by conflating them, consumers can end up disappointed by their tech, and companies can find themselves spending money on “solutions” that don’t really solve any problems.

Here’s the difference between the two—and why it pays to know which is which:

When customers ask for smart devices, vendors sell them IoT devices.

IoT devices work by continually monitoring something—heart rate, speed, temperature, etc.—and communicating this information to a data collection service, typically running in the cloud.

That data can then be used to optimize operational aspects of the asset (a human trying to sleep better, for instance). But IoT devices don’t make decisions for you. They monitor and collect data. That’s all. They depend on separate services to generate actionable information.

Embedded intelligence is different. When we give an object embedded intelligence, we’re giving it the ability to change something or make a decision in real time, at the source of the data collection. Sleep trackers use IoT tech because they simply monitor how well or how poorly you sleep. They do not exhibit embedded intelligence as they don’t adjust the firmness of your mattress or gently fluff your pillow in response to the data they collect.

Making local decisions to create a “smart” device is the purview of automation and has a long history. For example, you can find a very simple form of embedded intelligence in the governor on a steam engine. As the engine speeds up, the governor begins to spin. As the centrifugal force increases, the flywheels on the governor rise and limit the amount of steam that can reach the cylinders in the engine, keeping it from overheating. As the engine slows, the flywheel comes down, allowing more steam to reach the cylinders. Essentially, the governor is reacting as the situation changes, making a “decision” about how much steam to allow into the cylinders.

Today, embedded intelligence is much more capable as the cost of sensors has gone down and processing capability has gone up significantly, making it possible to collect and analyze vasts amounts of information, but the principle is the same. A change or decision is being made on the fly, at the source of the data.

The value difference between IoT and embedded intelligence:

Let’s take an autonomous car as an example. The ability to act on the sensor data to facilitate autonomous driving is an example of embedded intelligence, and it creates a completely new product category. Sending this same sensor data to an IoT data collection service would only increase the cost of the product, and provide no value-add. However, if we send acoustic, suspension, or engine performance data that holds information about the mechanical performance of the car and its components, an IoT service could provide value-add through prognostics to generate predictive maintenance alerts.

Embedded intelligence is about acting on the here-and-now to create new product categories. IoT is about data collection and analysis to discover the behavior and characteristics of the device, and the collective intelligence that can be generated when looking at thousands or millions of these devices in unison.

In short, IoT approaches are vast undertakings that require careful planning and follow-through. Value will not be created until the effort crosses a threshold of sufficient scale and complexity.

Going back to the sleep tracker, the current state of that effort hasn’t crossed that complexity threshold yet, and thus no value is being created.

Embedded intelligence and IoT are both data-driven endeavors—but they work at very different scales. Both require deep insights in the mathematical models that accurately describe the operations of the device, but the goal of embedded intelligence is to condense this insight into immediate action at the device, whereas IoT needs to condense that insight across the collective of devices.

Ultimately, conflating IoT and embedded intelligence isn't just inaccurate—it's costly, time-consuming, and not likely to meet your customers' expectations.