One of the many delayed Apple Intelligence features is known as App Intents, and we’re starting to see evidence that taking an extremely cautious approach to the rollout may be no bad thing. Before the comments catch fire, I should stress that I’m most assuredly not giving Apple a free pass on the slow rollout of new Siri capabilities in general. There are a great many capabilities which should very definitely have been launched years ago. Indeed, I’ve argued that the delay is now so embarrassing that Apple should probably allow us to choose our own chatbot to stand in for Siri, and the majority of you agreed with me. But when it comes to AI agents, taking things slowly may be the right approach … The power of AI agents It’s now almost a decade since I first described what I would like Siri to be able to do. I gave the example of the instruction, “arrange lunch with Sam next week.” Siri knows who Sam is, so that bit’s fine. It has access to my calendar, so knows when I have free lunch slots. Next, it needs to know when Sam has free lunch slots. That requires Sam to have granted me access to her calendar at a simple busy/free level. My iPhone checks Sam’s iCloud calendar for free lunch slots and matches them with mine. It finds we’re both free on Wednesday so schedules the lunch. And it could do even more. My iPhone could easily note my favourite eateries, and Sam’s, and find one we both like. It could then go online to the restaurant’s reservation system to make the booking. Fast-forward to today and Google has introduced the restaurant booking part of this. We’ve also seen this kind of AI agent capability rolled out on a number of other platforms, especially in the enterprise sector. Apple’s own offerings, meantime, are nowhere to be seen. App Intents Apple’s approach is to introduce this kind of capability through a mechanism known as App Intents. The starting point is to allow Siri to pull information from apps, something the company demonstrated back in 2022. We’re currently expecting to have to wait until next spring before this is launched. But the bigger goal here is to give iPhones AI agent capabilities, so that they can carry out more involved tasks on our behalf. But AI agents are not yet very reliable The problem, not just for Apple but for the entire tech industry, is that AI agents are not currently very reliable. Google notably goes as far as entering all of the information in the restaurant booking form before requiring the user to double check it and submit the booking. AI Mode will directly link to the booking page so you can finalize the reservation. TNW’s Calum Chase pointed to a couple of rather worrying statistics: More than half of companies (51%) have deployed AI agents 80% of them have disclosed that their AI agents have made rogue decisions Worse, he says many businesses are being very gung-ho about using these agents without proper oversight. Giving them autonomy with minimal checks is like handing the company keys to an intoxicated graduate. They are enthusiastic, intelligent, and malleable, but also erratic and in need of supervision. And yet, what large enterprises are failing to recognise is that this is exactly what they are doing. AI agents are being “seamlessly” plugged into operations with little more than a demo and a disclaimer. In this one area, a slow rollout may be smart Even the more basic usage of App Intents, namely pulling information from an app and presenting it to the user, could be fraught with peril if that information isn’t accurate and users rely on it. Carrying out tasks on our behalf becomes exponentially more dangerous. With individual users perhaps even more likely than businesses to just assume that Apple Intelligence will do what they have asked, maybe the company taking more time to do more testing in this one area is no bad thing. Photo by Mohamed Nohassi on Unsplash