I turned off Siri on the Mac years ago and never looked back. Similarly, I found Apple Intelligence so fruitless I never engage with it. But the new Siri AI coming to macOS 27 Golden Gate has at least got me slightly rethinking things.

I’m still early in testing Siri AI, as I’ve only had access to it in the macOS 27 developer beta for little more than 24 hours. It’s also in an early preview state on the dev beta, so there should be lots of runway for improvements before it releases later this year. I don’t even know if it’s done indexing my files and folders on our review unit M5 MacBook Air and M5 Max MacBook Pro. Unlike on the iOS 27 dev beta, there’s no “indexing in progress” box in the settings page. I asked Siri if it could tell me, but it told me to click a button in Settings that isn’t there.

My colleagues got a headstart testing Siri AI on the iPhone and Apple Watch, and getting a read on its general vibe, and they’ve so far had some positive feedback from using it. My feelings are a little more mixed.

When I sit down at a laptop I don’t need a voice assistant for searching things I’m randomly curious about or checking the weather like I would on a phone; I can do that faster and more accurately with keyboard and mouse. So I tried to think of ways to let Siri AI help me on macOS — things that might actually be useful to me in my everyday work.

I’d be happy to automate some of the time-consuming benchmarking I do when reviewing laptops, but although Siri AI can launch apps, it can’t take actions inside them (not that Apple ever claimed it could). I then tried to see if vibe coding a couple Shortcuts could get me there instead. This isn’t a Siri AI feature, but it is a new part of Apple Intelligence. I asked Shortcuts to run a test in either Geekbench or Cinebench, capture the results in a screenshot, wait a few minutes, and repeat the process two more times. But the resulting automations couldn’t actually run the tests either. Apple Intelligence made a shortcut that opened Geekbench and took screenshots (but forgot about actually running the benchmark), and it made a Cinebench shortcut that had “Wait for you to run the test” as an actual step. Maybe if developers continue expanding App Intents this could one day work.

So if Siri can’t help me run my benchmarks, maybe it can at least help me be a little faster in logging the data. In my normal workflow, I run each benchmark three times, taking screenshots as I go, and later average out the results before cataloging them in a spreadsheet. Apple’s WWDC keynote showed someone using Ask Siri in Spotlight to analyze data in local files. So I tried selecting batches of those screenshots in Finder and asking Siri to calculate the average scores for me. It worked pretty well — most of the time.

It was smart enough to distinguish single-core CPU scores from multicore CPU scores and GPU scores, average the test results, and arrange them in easy-to-read tables. But it could get thrown off if I included screenshots of too many different types of tests, especially if I mixed ones with synthetic score results (Geekbench, PugetBench, etc.) and time-based results (Blender render tests and our 4K video export test). And it sometimes got thrown off by the CPU rankings data that’s visible in Cinebench screenshots. Ideally, I’d be able to have Siri AI accurately calculate the 15 or so averages from my dozens of screenshots all at once — that would save me some serious time. But for now, it can at best only help me a little bit. And unless it gets better I’m still inclined to continue doing it all myself, especially since Siri messed up the numbers a couple times by pulling the wrong data.

So far, Siri AI seems a lot more capable within Apple’s ecosystem than it is outside of it, even for apps and files that are already on my Mac but in non-Apple apps. When I asked Siri to find my pictures of cats or babies, it pulled up results from Apple’s Photos and Messages apps. This could be enough for plenty of people, but not for me. Most of my messaging is done in Signal, and photos from my phone are uploaded to Google Photos, not iCloud. Siri also missed the thousands of images I have in my Lightroom Classic catalog, even though the files are stored locally in the Pictures folder and I kept asking it to access them directly. It’s possible those files haven’t been indexed yet, but I have no way to tell.

For now, I’m getting similar vibes to when I tested Copilot Vision last year. Like Copilot Vision, you can use Siri’s Visual Intelligence to ask questions about things on your screen. And like Copilot, it’s limited. I asked Siri to evaluate benchmark results on a spreadsheet in Google Sheets, but it can’t see all the data if it’s not visible onscreen all at once. I could get it to see the whole spreadsheet by downloading it as an Excel file and pointing Siri at it in Finder, but when I asked for the laptop with the highest single-core Geekbench score it gave me multicore data. Not great.

I opened Siri while running Lightroom Classic, on a black-and-white photo from my Ricoh GR IV Monochrome review, and asked Siri how to make it look more like a shot from street photographer Alan Schaller. Siri offered specific value adjustments for exposure, contrast, and so on, and adjusting those values gave me a decent result. Unfortunately, when I asked Siri to judge the result, it went sycophant on me, saying I’d nailed the look and achieved an “almost timeless feel,” which is the type of behavior Apple says it’s not supposed to show. (I thought we were past this.)

Then I uploaded a classic Garry Winograd photo and asked how to change my Lightroom settings to match that photo; Siri recommended I set the exposure to the value it was already at. So, some hits and some misses.

It’s still very early days for Siri AI, and much can change between now and the final release. But what’s clear already is that the experience is likely to be very different on an iPhone, where so much of your data is held inside Apple’s apps, and on a Mac, where you’re likely to be bouncing between all kinds of apps and ecosystems in ways that limit what Siri can do. Even so, it’s faint praise, but this is still the most useful and helpful Siri has been. It’s baby’s first real AI steps for Apple.