For many years, the Internet of Things was treated as one of the most promising technologies behind the Fourth Industrial Revolution. Every conference about Industry 4.0 seemed to include discussions about connected factories, smart cities, intelligent buildings, precision agriculture, remote healthcare, and autonomous supply chains.
IoT was the blue-eyed technology of the industrial world.
Companies installed sensors. Governments launched smart city programmes. Manufacturers connected machines. Universities created IoT courses. Technology providers built platforms capable of collecting data from thousands of devices.
Then ChatGPT arrived.
The public release of practical generative AI in late 2022 created what many people described as the “ChatGPT moment.” Almost overnight, technology conversations shifted towards large language models, copilots, AI assistants, generative content, autonomous agents, and increasingly powerful reasoning models.
Every major technology company wanted its own large language model. Every organisation wanted an AI strategy. Every product presentation suddenly contained the word “AI.”
IoT was quietly pushed into the back seat.
It appeared as though AI had killed IoT.
Yet by the beginning of 2026, organisations began discovering an uncomfortable truth. AI could generate impressive answers, summarise documents, write reports, and create software code, but it could not understand what was happening inside a factory, farm, building, city, hospital, or supply chain unless those operations were continuously producing trustworthy data.
AI had intelligence, but it lacked eyes and ears.
That realisation is bringing IoT back from the grave.
AI Entered the Business, but It Could Not See the Operations
The first wave of generative AI focused heavily on information that already existed in documents, databases, emails, websites, and enterprise systems.
This worked well for knowledge-based tasks. AI could help employees search company policies, prepare proposals, analyse written reports, respond to customer questions, and summarise meetings.
The problem became visible when organisations attempted to use AI for physical operations.
A factory manager might ask an AI system why production output fell during the previous shift. The model could describe possible causes, but it could not identify the actual cause without current information from machines, production lines, energy meters, environmental sensors, and maintenance systems.
A plantation owner might ask whether crops needed more water. The AI could explain irrigation principles, but without soil moisture, rainfall, temperature, humidity, and water-flow data, it would still be making an educated guess.
A building manager might ask why electricity consumption increased. Unless the organisation had connected meters, equipment monitoring, occupancy information, and environmental data, the AI would have no operational evidence.
This is the weakness that many AI programmes are now confronting. AI projects often struggle when organisational data is fragmented, inconsistent, outdated, or disconnected from real operations. Recent enterprise discussions increasingly stress that AI requires a strong operational and data foundation before it can produce dependable business outcomes.
The problem is not always the intelligence of the model.
The problem is that the organisation remains operationally blind.
IoT Was Not Dead. It Was Waiting for AI to Need It
IoT never disappeared completely. Sensors continued operating in factories, cities, farms, hospitals, vehicles, and buildings. Connected devices continued sending temperature, vibration, pressure, location, energy, movement, air quality, and equipment-status information.
What disappeared was much of the attention.
Generative AI captured the imagination of business leaders because its capabilities were immediately visible. Anyone could type a question and receive an answer within seconds. IoT remained behind the scenes, quietly performing the less glamorous work of connecting devices and collecting data.
The relationship between the two technologies is now becoming clearer.
IoT captures what is happening.
AI interprets why it is happening.
IoT provides the operational evidence.
AI identifies patterns, predicts risks, recommends actions, and explains the situation in language that people can understand.
Neither technology can reach its full potential alone. AI without IoT may understand documents but remain blind to physical operations. IoT without AI may collect millions of data points without helping people decide what to do next.
AIoT brings both sides together.
Three Developments Are Giving IoT a New Life
The return of IoT is not a repeat of the earlier sensor-and-dashboard era. Its new role is closely tied to AI and automated decision-making.
1. AI Is Moving to Edge IoT Devices
AI models are increasingly being deployed on cameras, gateways, industrial controllers, vehicles, drones, wearables, and other connected equipment.
Instead of sending every piece of information to the cloud, an edge device can analyse data close to where it is generated. A camera can detect unusual behaviour without continuously streaming video. A machine gateway can identify abnormal vibration before equipment fails. An agriculture device can detect crop conditions and trigger an immediate response.
Processing information at the edge can reduce response time, lower bandwidth use, improve privacy, and allow some operations to continue even when cloud connectivity is unavailable. Industry research also identifies performance, compliance, privacy, security, and lower operating costs as major reasons for applying AI directly to IoT data sources.
The IoT device is no longer merely reporting what it senses.
It is beginning to interpret what it senses.
2. AI Is Being Added to IoT Platforms
The second development is happening at the cloud and middleware layer.
IoT platforms have traditionally handled device connections, data collection, storage, dashboards, alerts, rules, and device management. With AI, these platforms can analyse information from many devices, locations, and systems at the same time.
An AI-enabled IoT platform like Favoriot can identify recurring equipment problems, detect abnormal energy consumption, compare performance across several facilities, explain unusual events, and recommend where an operations team should investigate.
This shifts the platform from being a data repository into a decision-support layer.
The dashboard may show that a machine is overheating. The AI layer can examine temperature, vibration, production load, maintenance history, and surrounding conditions before suggesting a likely reason.
The organisation receives more than visibility.
It receives context.
3. MCP Is Connecting LLMs with Real-World IoT Data
Model Context Protocol, or MCP, is opening another important path.
MCP provides a standard method for AI models and agents to access external tools and data sources. A 2026 study demonstrated how MCP could allow large language models to interact with IoT data spaces, while earlier research proposed edge-based MCP servers that connect language models with sensors and physical devices.
This means an IoT platform with an MCP server could allow organisations to use their preferred AI model to ask operational questions in natural language.
A manager could ask:
“Which cold-room location showed abnormal temperature changes today?”
“Why did energy consumption increase at Building B?”
“Which water pumps are showing signs of possible failure?”
“What happened at the production line before the alarm was triggered?”
The language model does not need to contain the operational data inside its original training. Through MCP, it can request current and authorised information from the IoT platform, analyse it, and present an understandable response.
MCP is becoming a bridge between conversational AI and the physical world.
The New AI Strategy Begins with Connected Operations
Organisations should stop treating IoT and AI as separate technology programmes.
A serious AI plan must begin by identifying the operational questions that the organisation wants answered. The next step is to determine whether the necessary data exists, whether it is trustworthy, and whether it is available continuously.
Where the data does not exist, sensors and connected devices may be needed.
Where the data exists but remains trapped inside separate systems, an IoT platform can provide a common operational layer.
Where teams receive too many readings but too little meaning, AI can help explain patterns and recommend responses.
The practical sequence is straightforward:
- Identify the decisions that need better operational evidence.
- Find the blind spots preventing those decisions.
- Connect the required assets, equipment, environments, and processes.
- Collect continuous and trustworthy data through an IoT platform.
- Apply edge AI, cloud AI, rules, analytics, or language models where they create clear operational value.
- Link insights to alerts, workflows, and accountable actions.
Buying an AI model before fixing operational visibility is like hiring a brilliant consultant and refusing to provide access to the factory floor.
The consultant may sound intelligent, but the recommendations will still be based on assumptions.
IoT Has Returned with a Bigger Purpose
IoT’s first era was about connecting things.
Its second era is about helping AI understand the physical world.
The technology that appeared to have been buried by generative AI is now becoming part of the foundation that enterprise AI needs. Sensors provide awareness. IoT platforms provide continuity and context. Edge computing provides immediate local responses. AI provides interpretation, prediction, and decision support. MCP provides a standard doorway through which language models can interact with real operational systems.
The next generation of successful AI projects will not be judged by how impressive their chat interfaces appear. They will be judged by whether they can see what is happening, explain why it is happening, and help people act before a small issue becomes a major problem.
IoT has not returned merely as an old technology receiving another round of attention.
It has returned as AIoT, carrying a clearer and more urgent purpose.
The organisations that connect their operations today will give their AI systems the trustworthy evidence needed to make better decisions tomorrow.
The question is no longer whether IoT is still relevant.
The real question is whether the organisation’s AI can see, hear, and understand what is happening in the real world.
Is the organisation ready?
Leave a Reply