IoT devices are ideal sources for real-world data to feed enterprise AI systems. The IoT sector has begun to take advantage of this to deliver more business benefits.
IoT devices and systems are ideal sources for real-world data to feed enterprise AI systems to drive efficiencies in operations. Getty
A colleague recently asked, “Is IoT dead? The Internet of Things was a big deal ten years ago, but it’s not headline news anymore.” My short answer was, “No, IoT is not dead, but you’re right about the novelty wearing off.” IoT technologies are evolving beyond the hype to connect real-world industrial and consumer “things” with AI-based, high-ROI applications. Like all successful enabling technologies, IoT becomes less visible but more valuable as it matures. This article is my long answer to my colleague’s question. It explains how the combination of three tech trends—platform-based devices, AI’s voracious appetite for real-world data and solution-focused product development—elevate IoT from novelty to profitability, enabling a wave of killer apps.
Killer Apps In Industrial Verticals
First, some context. A “killer app” is an application so valuable that it makes underlying products and technologies indispensable. It also disrupts the status quo, like Microsoft Office transformed office work, Uber upended the taxi business and media streaming services upstaged the music and TV industries. IoT is well-situated to foster killer apps. At its core, IoT is a set of enabling technologies that connect all kinds of electronic devices to business and consumer applications. A killer IoT application uses connected devices to deliver industry-changing benefits, not just incremental improvements or product-specific features. IoT powers diverse applications across multiple vertical industries, but none has achieved “killer” status yet. For example, “smart home” consumer applications offer many convenience and efficiency improvements, and successful products such as smart thermostats, door locks and cameras are big sellers for specific use cases. However, while intelligent product features are compelling, they fall short of “indispensable.” Killer apps deliver higher-level benefits, such as autonomously optimizing whole-home electricity consumption or enhancing safety and security through situational awareness across all sensors.
AI And IoT
AI is the North Star guiding IT investments as enterprises discover unprecedented opportunities to enhance operational efficiency, improve product quality and create actionable business insights. For enterprises with physical infrastructure dependencies, these benefits depend on collecting and analyzing vast amounts of data about company operations and product performance. That’s where IoT comes in. IoT devices provide the real-world, real-time operations data that drives business transformation in the AI era. Connected devices are essential sources of truth for AI training and inference, so using their data to automate and optimize physical operations is among the first IoT killer apps. Killer IoT apps can emerge in any vertical industry that requires synergy between real-world “things” and AI-based applications. Manufacturing, energy, agriculture, transportation, construction, retail, healthcare, smart buildings and smart homes are all good candidates. AI-enabled applications in these verticals need vast amounts of operational data, and IoT demand is skyrocketing . . . yet killer apps are slow to emerge.
Removing IoT Deployment Barriers
Enterprises have deployed IoT in various forms for decades. Still, three barriers have delayed the emergence of killer apps. Extracting business value from large streams of diverse device data is difficult and costly. Device development doesn’t scale to meet demand. Most IoT-oriented companies deliver enabling technologies (devices and data services) without directly addressing customers’ transformational business opportunities. To put it another way, they overfocus on connectivity gadgetry and underemphasize the bigger picture. The good news is that three trends are now flattening these barriers. At the same time it creates growing demand for operations data, AI greatly simplifies the processing and contextualization of machine-generated data streams. IoT device development is evolving from custom mashups to platform-based approaches. Progressive, IoT-savvy companies are upleveling product strategies from device connectivity (“Let’s connect everything!”) to delivering killer apps to drive business transformation (“Let’s help customers improve operational efficiency, product quality and business insights!”). These three trends combine to enable killer IoT-enabled apps. Let’s dig into the details.
Trend 1: AI Changes The Game
AI accelerates IoT in three ways. First, enterprises are scrambling to optimize and automate business processes with AI. These new deployments are hungry for real-time (or nearly real-time) data from industrial processes and physical assets, so demand for operations technology is exploding. Automation and optimization “killer apps” will define the next decade of industrial IoT development and create insatiable demands for IoT-sensed data. Second, multimodal AI models trained with diverse data types, sources and formats can reduce the need for customized middleware that reformats, remodels and algorithmically analyzes operational data. IoT data is notoriously diverse, generated by many different device types, brands, architectures, ecosystems and vintages. Much of it is unstructured—not organized in any specific manner. Writing and maintaining procedural middleware to parse, organize and combine various incoming IoT data streams is often difficult and costly, and it’s a big reason many IoT projects fail. But multimodal AI is surprisingly good at extracting information from unstructured data pipelines. It can also combine information across pipelines, generating enterprise-wide insights spanning independent operational domains. Third, even small, battery-powered IoT devices can now use AI inference to convert raw data to actionable events at data collection points. On-device AI reduces cloud computing costs, cuts response times and addresses many data analytics problems more efficiently than procedural algorithms. Examples include recognizing abnormalities in sensor data streams, visually inspecting manufactured parts, recognizing people and activities and emulating human-like capabilities such as digitizing analog gauge readings. The commonality is that local on-device intelligence extends the reach of business transformation to include new types of operational processes. Summary: Rapid enterprise AI adoption increases demand for operational data from IoT sources; multimodal AI simplifies ingesting, processing and preparing data for use by AI-enabled business processes; and on-device AI makes IoT devices smarter, more autonomous and more capable of digitizing enterprise operational infrastructure at a lower cost.
Trend 2: Platform-Based, Software-Defined IoT Devices
Strong demand for process automation creates lucrative opportunities for IoT-based products. However, IoT device development is a narrow bottleneck in the product supply chain, mainly because embedded hardware and software require product-specific customization. Bespoke, one-off device development is complicated, slow, costly and risky.
Customization also drives up long-term support obligations, creating significant long-term technical debt. Indeed, customization is a big reason that only about 20% to 30% of enterprise IoT projects deliver on expectations, and this disappointing fact hasn’t improved much in the past five years. Yet IoT devices require customization because embedding electronics into “things” imposes practical constraints on hardware and software systems. Traditional small, low-power electronic devices have tiny processors, minimal memory and manually tailored OSes, plus they require specialized, low-productivity application programming techniques. Constrained devices were okay for simple, single-function products. However, modern IoT devices need multiple network stacks, enterprise-grade security, cloud-native services, over-the-air updates, high-level programming languages and other advanced features. Integrating this new system-level functionality into one-off custom products is unsustainably difficult and adds no product differentiation. If developers could use IoT platforms as-is, IoT application development would be comparable to writing apps for mobile devices and PCs. In other words, developers would build applications, not operating systems.
Technical Feasibility of Platform-Based IoT
Let’s look at the technical feasibility of this approach. Many leading semiconductor companies such as NXP, Qualcomm, Silicon Labs, STMicroelectronics and TI offer high-performance, power-efficient, single-chip MCU- and MPU-based SoCs with sufficient compute power to support a wide variety of IoT apps. These companies also provide developers with all required system software, tools and prototyping boards. Although these powerful, application-independent platforms are more expensive than stripped-down embedded chips, the benefits of faster application development, better security, higher product quality, cloud-native APIs, containerized applications and built-in long-term maintenance (now a regulatory requirement) more than offset the added cost. Platform-based IoT allows developers to skip systems engineering and security work that adds no customer value and start building applications immediately. The result is better applications, better security, faster time to market, lower risk, lower development costs and much lower long-term maintenance costs.
At this point, my embedded hardware engineering colleagues roll their eyes because squeezing IoT devices to fit within real-world products has always required unique form factors, tight power constraints and unique circuit designs. That’s still correct, but platform-based IoT simplifies the process. Modern IoT SoCs pre-integrate all required function blocks including networking, which makes product-specific hardware design easier. System-on-module platforms further simplify customization by limiting hardware development to power supplies, I/O interfaces and physical packaging. For example, the Arduino Pro products offer complete compute subsystems on very small modules, some measuring less than one square inch. Container technologies have the potential to take platform-based IoT to the next level by delivering applications much like apps on phones. Docker/Kubernetes, WebAssembly and other container frameworks are now hot topics among IoT architects. There are many complications, but isomorphic, efficient, platform-independent containers are on the horizon, and product companies should monitor this tech trend carefully. I’ll cover this topic in a future article.
Bottom line: The device development bottleneck is disappearing as developers use fully integrated platforms as-is from semiconductor vendors and system software suppliers, complete with OS, security, networking and development support. In this context, IoT companies should avoid developing undifferentiated system-level software unless there is no alternative. Transitioning from DIY devices to off-the-shelf platforms reduces time-to-market, improves application software quality, increases security and reduces risk. Application containers are the next step, further abstracting app development from embedded systems.
Trend 3: Product-Centric IoT
AI-enabled applications using platform-based, software-defined IoT devices open the floodgates for a torrent of connected industrial products that deliver significant and measurable enterprise business value. As IoT technologies mature, industrial customers can migrate from DIY projects, custom devices and hard-coded middleware to off-the-shelf IoT products from suppliers with a track record of delivering well-defined business benefits. Forward-looking industrial IoT suppliers embrace the shift from technology-centric IoT to product-centric IoT, adopt customer-focused mindsets and avoid treating IoT ingredients as deliverables. These companies transparently integrate mature, horizontal IoT connectivity features into vertical industry products that directly address customer business situations.
The shift to product-centric IoT is difficult for companies that make the classic mistake of confusing ingredients with products. IoT is an ingredient—a set of enabling technologies—not a product. For the past decade, immature IoT technologies have forced many enterprises to source hardware and software components from various ingredient suppliers, develop IoT solutions in-house or with costly contractors and struggle with complicated OT-IT integration middleware. The success rate for these projects was terrible—around 20%, by most estimates. Product-centric IoT suppliers embrace AI-enabled, software-defined device platforms, resulting in a step-function change in scalability. This change differentiates IoT ingredient (technology) suppliers from product (application) suppliers. Today, most IoT-oriented companies are ingredient suppliers—although some won’t admit it. Each ingredient company must adapt as its customer base shifts upstream from enterprises to product-centric IoT companies.
Product-Centric IoT In Action: Samsara
Opportunities for product-centric IoT companies abound.
A few are already in leadership positions in their industries, sharply focused on solving customer problems and poised to deliver killer apps. Samsara (NYSE: IOT), an IoT company in the physical operations industry, is a good example. Samsara’s customers span construction, transportation, logistics, food distribution, chemicals, energy, municipalities and other industries. These enterprises have one thing in common: the need to digitize physical operations in industrial environments across large geographic areas. Samsara addresses this market with rugged, GPS-enabled, cellular-connected gateways installed in trucks, warehouses and industrial equipment. The gateways connect with telematics, cameras, refrigeration units, tracking devices and other equipment via native plug-and-play interfaces. On the other end of the cellular link, Samsara’s Connected Operations Cloud collects, manages, analyzes and generates insights from more than 9 trillion data points per year. Customers interact directly with the cloud-based applications via smartphones and tablets, and dozens of pre-integrated connectors such as Microsoft Power BI visualization and Kafka (now in beta) simplify connections with other business systems. Samsara emphasizes ease of installation. A mechanic bolts a gateway into a truck, plugs in the cameras and telematics and the assets appear on the Operations Cloud. The company does not need to discuss hardware, firmware, networks or other technical issues with customers. Instead, customer interaction focuses on high-value solutions that improve fleet management, workflow automation, safety, regulatory compliance, employee performance and equipment uptime. These solutions are increasingly AI-enabled. For instance, cameras can detect driver inattention, drowsiness, lane departure and potential collisions. Samsara’s integrated solution is a killer app because the company uses AI to change the game, has no device bottleneck and focuses on customer-facing products rather than device infrastructure. In the future, I expect the company to double down on AI-enabled applications, adding valuable insights from the massive amounts of data that flow into the platform. I’ll cover the details in an upcoming article.
Turbocharging IoT Growth
Three technology trends combine to create a virtuous cycle that enables the first wave of killer IoT apps. AI technologies enhance IoT applications, middleware and devices. Off-the-shelf IoT platforms streamline device development while improving security and quality. IoT product companies evolve beyond devices and connectivity to provide business-focused applications that optimize enterprise operations and deliver new business insights. Development of AI technologies, IoT platforms and new IoT products work together in a virtuous cycle.
Enterprise operations automation generates ROI that increases demand for IoT products, which increases investment in IoT platforms and AI technologies, enabling IoT product makers to deliver new generations of high-value, AI-powered applications—and the turbine just keeps spinning faster. The most successful products become killer apps: essential pillars of digital transformation for industries with physical assets. Future articles in this series cover product companies driving digital transformation in specific verticals. As I prepare this work, I’m analyzing dozens of companies, including Samsara, Honeywell, Johnson Controls, Schneider, Siemens, Infinite Uptime, ThinkIO, John Deere, Agtonomy and Sonatus. I’ll also cover innovative silicon and software platform suppliers that accelerate the IoT growth turbine.
Moor Insights & Strategy provides or has provided paid services to technology companies, like all tech industry research and analyst firms. These services include research, analysis, advising, consulting, benchmarking, acquisition matchmaking and video and speaking sponsorships. Of the companies mentioned in this article, Moor Insights & Strategy currently has (or has had) a paid business relationship with Honeywell, NXP, Qualcomm and Schneider Electric.
Bill Curtis
Following Editorial Standards
Print Reprints & Permissions
Bill Curtis,Contributor