On-Device AI Processing Could Change Mobile Commerce

Artificial intelligence (AI) is being embedded directly into devices, potentially boosting app speeds dramatically.

As Apple reportedly gears up to introduce AI-powered features on the iPhone that rely on on-device processing rather than cloud servers, experts are considering the potential benefits of this approach for the future of mobile commerce.

For mobile commerce applications, where speed and convenience are crucial, on-device AI could enhance the user experience and drive increased adoption of AI-powered shopping and payment features.  

“This type of onboard processing means that people can interact with products or services quickly in situations where connectivity is not available,” Tony Fernandes, CEO of UEGroup, told PYMNTS. “It will provide enough speed and power to provide really amazing customer experiences quickly without the delays that frustrate people and create abandonment.”

AI Onboard

Apple’s upcoming AI features in iOS 18, including a new version of Siri, will operate solely on the device, bypassing cloud servers. Bloomberg reported Sunday (April 14) that the first batch of enhancements set for release on June 10 will rely entirely on local processing, eliminating the need for cloud-based computation in Apple’s as-yet-unrevealed large language model.

Samsung, a major competitor of Apple, has already implemented on-device AI in its Galaxy S24 series.

This type of AI is also used in various forms on many computers. For example, the Stable Diffusion model, a text-to-image AI model, is typically set up to run locally on users’ machines. However, Kirk Sigmon, an AI specialist at intellectual property protection firm Banner & Witcoff, told PYMNTS that it has limitations, such as its slow performance on older computers. 

“On-device AI stands to increase user trust in AI implementations, particularly those involving personal data,” Sigmon said. “At the same time, it’s going to raise the bar for on-device processing capabilities — after all, many cooler implementations of AI, e.g., generative AI, can’t run on lower-powered and cheaper hardware.” 

Onboard processors allow AI computation to be done offline. Fernandes noted that some real-world uses include being on a flight without connectivity or in a remote area with no cell coverage.  

Another significant advantage of processing AI on a device is speed. Even with 5G mobile phone service, or with upcoming 6G, there is latency — or lag time — associated with transmission on the network, processing data on a server, and then transmitting it back. 

“These same principles apply to mobile phones in general, which is why they still have powerful processors and local storage,” he added. “Apple’s strategy makes sense for that reason and because it will give them control over what gets optimized on the chip. Makers of general AI processors have to take into account a lot of markets that may not matter to Apple.” 

One example of this new type of onboard AI is software from BlueSkeye AI, which has built an AI-driven mental health system that runs on people’s mobile phones. It uses machine learning to analyze face and voice data and measure medically relevant behavior, such as pain.

Michel Valstar, the co-founder and chief scientific officer at BlueSkeye AI, told PYMNTS that processing power and battery consumption are constraints.

“We make the most out of small machine learning models and modularize the AI into several components where the output of one serves as the input to the next,” Valstar said. “This reduces the amount of variables going into each next component at every step.

“This results in smaller neural networks using less computing and power. A nice side effect is that we need less training data for each component,” he added. 

Privacy Boost

Enhanced privacy and security are key benefits of processing data with AI directly on devices. Supporters believe that keeping data on the device makes it less likely to be intercepted or accessed by unauthorized people.  

However, it’s important to note that simply keeping data on the device doesn’t guarantee safety, Alan Bavosa, vice president of security products at Appdome, told PYMNTS.  

“An increase in sensitive data being stored on-device will most certainly attract even more sophisticated types of threats such as malware, malicious bots, trojans and spyware which specialize in targeting and exploiting such data due to the lucrative potential it offers,” he said.