Punnam Raju Manthena, Cofounder & CEO at Tekskills Inc. Partnering with clients across the globe in their digital transformation journeys.

getty

The Internet of Things and AI can be a great combo, but there is always the possibility an AI system can make an automated decision that has unintended consequences.​ For an example of this, look no further than IBM's Watson, which, according to STAT News, "often spit out erroneous cancer treatment advice."

In most AI failure scenarios that I've observed across industries, no human actually approves the faulty decision, but the damage is lasting. In worst-case scenarios like these, who is accountable when AI-powered IoT solutions make an incorrect, life-critical decision?

​How IoT And AI Work Together

Let's go over some of the basics. The Internet of Things (IoT) is the linking of your various devices—be it Wi-Fi devices, Bluetooth devices, computer peripherals, smart appliances, etc.—that collect and share data. The IoT doesn’t have any intelligence of its own, but it works on simple, predefined rules like switching on a light when you trigger a sensor.

When IoT devices are combined with AI—which can simulate human intelligence and think, reason, make decisions, etc.—they can process humongous amounts of data to form smart, automated systems that can analyze information, learn from patterns and make autonomous, real-time decisions without human interference. This heady combo, based on my own experience in this space, can improve efficiency, automate complex tasks and enable true predictive capabilities across industries.

Some popular applications that I've observed include predictive maintenance in manufacturing to foretell equipment failures and avoid costly downtime, continuous monitoring of patients and automated healthcare alerts, better irrigation via the analysis of soil conditions, wetness and crop variety and better driving decisions in transportation.

​Success Stories

One major AI and IoT success story comes from BMW. They leveraged cellular connection to transmit data to the cloud for data analysis. Using Microsoft's Azure AI Foundry and Azure OpenAI Service, they were able to deliver and analyze data 10 times faster. Husqvarna Group also leveraged Microsoft Azure solutions to increase their data deployment speed by 98% and reduce their infrastructure imaging costs by 50%.​

According to Gartner, about 40% of all enterprise apps may have task-specific AI agents by the end of 2026—a major increase from less than 5% at the end of 2025.

​Common Failures

Despite their benefits, IoT and AI can sometimes fail in tandem. This can stem from incorrect data, for example. You could also misinterpret a user's intent. Your system may not be correctly identifying and classifying entities. Your virtual assistant may not have the correct context of a conversation. Your training data may not be complete or fully representative of real-world scenarios. Your AI algorithm could have an inherent or developed bias. There may be constraints in your model's ability to understand human language. There could be IoT solution or sensor degradation, connectivity issues and hallucinations because of poor update procedures and more.

Whatever the reason, AI does fail. According to The Guardian, Zillow, an American tech real-estate marketplace company, lost $304 million and 2,000 jobs because the algorithm they used to value homes undervalued volatility and didn’t consider industry shifts during the pandemic.

​Bridging The Accountability Gap

AI failures like these can often be attributed to an accountability gap, which can stem from an algorithm gap that exists between the performance of an AI model in a governed environment and its actual performance in the real world. Sometimes, the accountability gap could be caused by an operator gap, the divide between AI's potential and the actual real-world performance with a human prompter. Or, the gap could be regulatory in nature. In my spaces, the fast, global deployment of AI and the sluggish, fragmented development of legal governance frameworks can expose organizations to legal, financial and reputational risk when AI fails.

​Despite this, AI is here to stay. About 96% of technologists believe agentic AI will see tremendous growth in 2026. In India, the government is planning on installing about 250 million prepaid smart consumer meters across India by March 2026. This initiative is expected to help electrical distribution agencies improve their billing, collection and network infrastructure—and reduce their losses.​

With all of this in mind, it's important to look at the gap issue from a real-life standpoint—not from the theoretical angle. Develop a clear strategy and invest in an AI tool that suits your needs. Better still, start with a prototype, then go for full production. Try to govern your AI systems from a legal and compliance standpoint. Start small and slowly increase your usage—after all, ROI is all-important, and you wouldn’t want to spend dollars without getting anything in return.

​Conclusion

If your AI isn’t working, it isn’t about capability. It's about direction, intent and approach. Strategic clarity, prototyping and solid governance aren’t just optional implementations. They're critical aspects that differentiate tangible value from directionless work.

Don't just comply with standards and trends. Trust your intent, your direction, your tools and, most importantly, your people. AI is not the advantage; the advantage is what you make out of it.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?