For those who have never heard the term, “Cheap and Cheerful,” it is often used in reference to devices and modules built for basic telematics. Many of these devices are designed to be left out in the field for as long as ten years. In order to meet the requirements and battery life, these devices are normally restricted on computational power consumption.
So, I asked Haroon K of Publicis Sapient about the situation. Publicis Sapient started off as an IT consultant company but, today, has a keen focus on the issues of transforming the enterprise. It does this with a comprehensive suite of services, some of which are for ideation and others representing specific department needs, with a particular emphasis on Digital Transformation. The company has 50 offices around the world and over 20,000 employees.
Given they are change agents, I thought Haroon could tell me the if, where, and when for cheap and cheerful telematics devices and modules
Two words of caution: First, this was a text conversation, and the “quotes” are not exact, but they reflect the intent of the content and, secondly, some ideas may have been moved to make their emphasis less hidden.
I posed these questions to Haroon:
Generative AI uses Large Language Models (LLM) to interface with human requirements. I work within the AI/IoT crowd (or AIoT). The AIoT companies work with much smaller vocabulary. Typically, the amount of data being sent is small and the features and functions are pretty narrow.
Given the difference of audience (Humans vs Machines) are AIoT developers a divergence from Generative AI and therefore not benefitting from Gen AI?
Secondly some of the chips and module makers are looking to add larger computational features to their hardware. That adds cost and I am not sure the market is looking for more “smarts” in the field. Does Generative AI have so much momentum that “cheap and cheerful” telematics is heading to a death spiral.
Haroon’s response and explanation: AIoT vs. Generative AI in Practice
AIoT is focused on narrow, purpose-built models optimized for low power, limited compute, and bandwidth-constrained environments.
Generative AI (Gen AI) operates at a different scale – mainly in the cloud or on powerful edge nodes, not on constrained field devices (sensors or actuators).
AIoT may benefit from Gen AI, especially at the powerful edge gateways (e.g., conversational interfaces, smart diagnostics).
Hardware makers are adding more compute, but many use cases still favor cheap, low-compute field devices with smarter gateways.
Edge gateways are becoming the key processing and decision points, running fast, lightweight ML models (called Tiny ML in some cases) – not Gen AI – to trigger actions based on real-time data. The model used is not a Generative AI model, but a fast, pre-trained anomaly detection or time-series forecasting model, optimized for the specific use case.
“Cheap and cheerful” telematics is not dead, but its value is shifting to how well it integrates with intelligent gateways and cloud systems.
Here’s a real example: Edge gateway responding to emergency sensor data cloud connect + industrial edge gateway
In oil & gas or water management networks, this setup monitors pressure and flow sensors across pipelines. If a sudden drop in pressure is detected – indicative of a leak or rupture – the edge gateway (e.g., running industrial edge platform) executes a local ML model to analyze pressure delta trends. The ML model confirms the anomaly via threshold and pattern models. Next, a trigger transmits to a valve actuator to isolate the section in real time, thereby reducing the latency.
This allows real-time, autonomous intervention without latency of cloud communication, which is critical for safety and operational continuity.
Conclusion:
I want to thank Haroon for using a case study from a Publicis Sapient deployment using AI/ML today. Given what I have heard and read from the hardware side of the house, it may be, in the future, that devices and modules end up with a SIC code that treats the additional computing capabilities as an extra service to be turned on in the module. However, it’s pretty clear that AIoT solutions are going to have a mind of their own for quite some time.
Edited by
Erik Linask
