Today (16 July) marks AI Appreciation Day, an annual event established in 2021 by entrepreneur and former Goldman Sachs analyst Nathan Ricks to encourage discussion around the responsible development and use of artificial intelligence.
It reflects the technology’s growing prominence – even if the subject of the celebration is itself ironically incapable of appreciating the gesture.
But beyond the celebration of artificial intelligence itself, the event highlights a broader shift taking place across industrial automation: the focus is moving from what AI can do to where, and how and why it should be deployed.
As manufacturers accelerate AI adoption, the next challenge is moving intelligence from the Cloud to places where decisions need to be made. Technology companies argue that the next phase of industrial AI will be defined not by ever larger Cloud-based models but by intelligent systems operating closer to machines, where they can deliver real-time decisions, greater resilience and stronger security.
For many industrial applications, that means moving AI from remote data centres to the network Edge.
“A Cloud-first architecture isn’t always the right answer,” said Dunstan Power, Director and Co-Founder of ByteSnap Design. “If your system depends on a reliable network connection, you have to consider what happens when bandwidth is limited, latency is unpredictable or there’s no connectivity at all. For many embedded applications, that’s simply not acceptable.”
Power said more industrial projects are deploying AI models directly on embedded hardware, allowing systems to continue operating even when Cloud connectivity is unavailable.
“We’re seeing more projects where running AI models directly on the device removes those constraints,” he said. “It allows systems to make decisions locally, respond in real time and continue operating when a Cloud connection isn’t available. That’s opening up applications in industrial, environmental and remote deployments that would otherwise be difficult or impractical.”
Machine vision systems inspecting fast-moving production lines are one example. Rather than sending images to a distant server for analysis, manufacturers can process data locally, enabling sub-second responses for quality inspection while reducing bandwidth requirements and improving resilience. Keeping AI models on the device can also help manufacturers meet data sovereignty requirements by ensuring sensitive operational data remains on site.
ByteSnap cautioned, however, against using AI for deterministic control of safety-critical systems. While machine learning can reliably identify defects or anomalies, the consultancy argues that conventional software should remain responsible for controlling industrial equipment and medical devices.
As AI becomes more deeply embedded within industrial operations, questions around governance and security are also becoming more prominent.
“AI has won the innovation race. Now it has to earn confidence,” said Chris Harris, EMEA Technical Director, Data and Application Security at Thales.
According to the company’s latest Digital Trust Index, 93% of IT and security leaders are already deploying or planning to deploy generative AI, yet only 23% of consumers trust organisations to use AI responsibly with their personal data.
Harris said businesses that prioritise governance alongside innovation would be better placed to build confidence in AI.
“The businesses that give equal weight to transparency, security and governance alongside innovation will be the ones that turn AI investment into lasting customer confidence – and, ultimately, appreciation.”
The research also suggests people remain more comfortable with AI supporting human decision-making than replacing it. While 63% of consumers are comfortable with AI supporting everyday online tasks, 77% said a company’s use of generative AI would not increase their trust.
The shift towards practical, outcome-focused AI is also shaping how businesses think about technology beyond the factory floor.
“The best AI is the AI you barely notice,” said Peter Bell, Vice President of EMEA Marketing at Twilio. “It’s the technology working quietly in the background to remove everyday friction, reduce small moments of stress, and help people get on with their day.”
Bell argued that the most successful AI applications would be those that quietly improve processes rather than draw attention to the technology itself.
“As AI becomes more embedded into the products and services we use every day, I think we’ll increasingly appreciate it not for the technology itself, but for the experiences it enables,” he said. “The future isn’t about replacing human interaction; it’s about removing unnecessary complexity so that when people do connect, those interactions are faster, more relevant, more personal and ultimately more human.”
