The big promise of AI agents is that they’ll be able to handle tasks for you — using their knowledge and understanding of you and what’s stored in your phone to suggest, predict and automate what you need, to ease the burden on you.
For the most part, the situations in which we’d use AI agents in our day-to-day lives have so far been largely hypothetical. But at Qualcomm’s Snapdragon Summit in Hawaii, I got a first-hand look at how we might use an agent to complete a routine task: uploading content to social media.
Using a prototype phone packing Qualcomm’s new Snapdragon 8 Elite Gen 5 chip, I asked the device, using my voice, to find all pictures of beaches stored in the Photos app. A large language model (LLM) running on the device picked up what I was saying and interacted with a vision model that classifies all the photos on the phone. It pulled up two options.
Don’t miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.
«I like the second option,» I told the agent. «Please post it to Instagram with the hashtag #lovethecolor.» Without my touching the screen, the agent opened the Instagram app on the phone and posted the photo as a Reel (which is what it had been preprogrammed to do). Again, the LLM kicked in, but rather than sending a command to the photo classifier, this time it sent the command to the Instagram API.
After posting, the agent also asked me if I’d like to check for new comments. I’m pretty sure this is what notifications are for, but this was just an example to show how proactive the agent was able to be.
In fact, the whole demo was just an example of how an agent could assist you in your daily phone business. In the US, most social platforms, including Instagram, don’t currently allow access to their APIs that would make this process possible. Qualcomm built the demo together with AI company ModelBest and is going to launch it in China on the popular social site Weibo.
After my demo, I’m not in a particular rush to engage the services of an agent to upload to Instagram for me. I appreciated the image classification tool most of all, since being able to describe a photo in your camera roll to post rather than having to scroll to find it was a definite time saver. But posting to Instagram is already a pretty slick and seamless process that I’m not sure warrants automating.
I’d also want the option to post to either Stories, Reels or the main grid, and give more complex instructions about editing, filters and captions before I’d be willing to hand over the reins to an agent.
For now, I’m happy to continue uploading to Instagram under my own steam, but I’m keen to see how agentic AI evolves to be able to handle more complex tasks and commands over time.
Qualcomm and many other tech companies are convinced that agents are gradually going to become the de facto way we interact with our technology. The jury’s still out for me, but I’ll keep an open mind.