Artificial intelligence is increasingly being integrated into IoT solutions and can even generate the code that enables devices to communicate with networks. In theory, AI is a surefire way to raise productivity in software engineering, but the reality may not be quite so cut-and-dried.
The world is becoming ever more connected, through IoT devices forecast to exceed 40 billion by 2034. Smart buildings are contributing to this growth, with a forecast market size of over $550 billion by 2033, a compound annual growth rate of more than 18%.
IoT systems monitor temperature, humidity, lighting, air quality, and more through sensors designed to improve building health and safety. They also support security measures through building access control and CCTV. During Covid-19, IoT even helped manage the flow of people through buildings with thermal-imaging cameras and occupancy-monitoring systems that supported social distancing.
To be more sustainable, smart buildings deploy energy-efficient IoT solutions. These provide insights that building and infrastructure managers can draw on to minimize power wastage and optimize energy systems. Predictive maintenance solutions collect data on the condition of equipment, reducing site visits and making better use of in-person appointments.
[View All](https://www.eetimes.com/category/sponsored-content/)
New technologies create new IoT opportunities. Ambient light harvesting, for example, can make building sensors self-powered. Over time, these sensors are likely to exchange data on, for example, light and power levels.
The case for AI****oT
Now, design engineering is integrating AI into IoT (AIoT) to take IoT further.
It makes sense. AI and IoT are natural bedfellows: IoT generates and captures data (often in large volume), while AI is ideally placed to analyze it. Such is AIoT’s potential, and research company Transforma Insights forecasts over 9 billion connections at the end of 2033—the culmination of a sixfold growth over 10 years.
One standout AIoT application is security. AI is adept at spotting patterns, so it can identify anomalies more readily than human oversight can. This can reduce threat detection from days down to minutes. As agentic AI progresses, humans could become even less involved if AI is configured to not only intercept issues but work to resolve them as well.
AI in software engineering
AI could potentially change software engineering, too, through faster code generation and problem solving. AI coding promises dramatic productivity gains and so is likely to appeal to companies striving to save costs, deliver more, and maintain a competitive edge.
That would, of course, impact jobs, skills development, and levels of control. However, while some organizations may be considering AI first, others are more cautious.
Gartner has predicted that 60% of organizations will have smaller software engineering teams by 2029 but stresses that the new “tiny teams” will constitute a restructuring and that AI will, in fact, drive demand for more software engineers. It warns organizations that cutting their junior roles because of AI would “hollow out their … software engineering talent pipeline.”
Some companies are reconsidering whether, and how, to adopt AI tools. Three factors contribute to this:
- The productivity anomaly: There is evidence that productivity expectations of AI coding are not being met. DX analysis revealed that, despite a 65% rise in AI adoption by engineering organizations, the speed of software delivery was up by just 8% on average, far below the anticipated threefold to tenfold productivity gain. Similarly, software and engineering intelligence company Faros discovered that while developers completed 21% more tasks with high AI adoption,code review times increased by 91% .
- Workflow bottlenecks: A higher volume of generated code exposes bottlenecks in workflow development processes elsewhere and reminds us that, while AI accentuates the (arguably unachievable) perfect processes, it also reveals process flaws. If coding is not a primary drain on software engineering time (and DX puts it at only about 16% of a developer’s day ), speeding it up will not move the output dial significantly.
- AI skepticism: Adding to the operational challenge and delivery shortfall of AI coding, there is some inherent mistrust of removing or reducing human involvement. One survey found that a resounding 96% of developers don’t fully trust AI-generated code, and 61% agree the code often looks correct but isn’t reliable.
Designing for the long term
Consider once more connected applications for smart buildings: They must be resilient across the complete end-to-end design chain of physical device, network, and cloud/core endpoint. Engineers and designers must ensure all elements work together seamlessly, reliably, optimally, and securely. Poor-quality code has no place in this complex reality and could, in fact, have serious consequences.
In IoT, the challenge goes beyond using AI to write software faster. Code must perform in the long term as technology, hardware, and circumstances evolve. Maintenance and security are paramount and cannot be overlooked or, in the worst case, compromised.
IoT solutions must be designed end-to-end with connectivity, security, resilience, and data management in mind. That is the only way to minimize in-life corrections that can impact customer and business outcomes and be costly both in time and money.
Keeping people in the loop matters, but not as a failsafe—people are fallible, too. The task is to apply engineering experience where it changes the outcome rather than position human review as the thing that catches what AI misses.
In practice, that means AI-generated code destined for production is reviewed by an engineer before it merges. Agents run first-line diagnostics on incidents and propose a remediation plan, but a person decides whether it is executed.
AI governance and control
AI will inevitably play an ever-increasing role in IoT at both the application and design level. While there is an appetite for this, there is also an understandable sense of caution.
Organizations must know what AI does, in all process integrations, understand why it does it, and have the necessary level of control to prevent deviations. The organizations that will succeed won’t be the ones adopting AI the fastest; they’ll be the ones that can say exactly where they draw the line and hold it.
See also:
EE Times Magazine – September 2026 The September 2026 edition of EE Times Magazine examines how smarter buildings combine ambient energy harvesting, sensing, AI, and connected systems to improve safety while protecting privacy.