Particularly beloved by coders, Claude is one of the best-known and most widely used AI chatbots around right now. One of its features is that it comes with Connectors: add-ons that let you link the prompt box to third-party services such as Spotify, Uber, and Slack.
Those Connectors include several Google apps, and I've been playing around with Claude's Gmail integration. Email is one of the biggest time sinks in my day, and if Claude could save me a few minutes here and there, it would be genuinely useful.
I was wary of letting an AI loose in my Gmail inbox, and there are reasons to be cautious, but Claude's analysis and actioning worked better than I expected. Here's how you can get started, and how the AI chatbot helped me.
Getting connected
There's obviously a privacy trade-off here: you have to be okay with Claude accessing your emails and seeing what you're up to.
Anthropic says personal data isn't used for marketing or to build up a profile of its users, though text may be used to train its AI models — if you're not happy with that arrangement you can disable it in Claude's settings.
There's also the very understandable worry that Claude will suddenly delete 100 emails behind your back, and it's another thing I was cautious about. However, I didn't come across any issues during my testing, and if you want Claude to take any actions on your behalf, you can force it to ask you for confirmation each time.
To connect to Gmail from the Claude web app, click the + (plus) button in the prompt box, then choose Connectors > Add connector and find Google's email platform.
You'll need to log into your Gmail account and confirm the connection, and you can then invoke the Gmail Connector by referring to "gmail" or "email" in your prompts.
You can also get some prompt ideas by clicking the From Gmail button under the prompt box. To begin with, I asked Claude to analyze my email organization methods: I like to be ordered and precise when it comes to email, and the AI was scarily good at spotting how I used labels, and the different buckets that I sorted my emails into (my inbox is a weird mix of work, friends, family, press releases, and general miscellany).
Claude also did a fine job of telling me which emails I often leave unread (newsletters, social media alerts, and promotions, mostly), and giving me a nudge about emails I haven't replied to. I also liked its recommendations for better optimizing my inbox, with intelligent suggestions for more labels and filters.
Your email inbox can say a lot about you, and Claude worked me out pretty fast. It can even do you a personal profile and interactive graphic based on your inbox: You can get an overview of the tones and styles you use most often. I'm "tersely efficient" and "low maintenance", so form an orderly queue, commissioning editors.
Managing emails
Scanning and summarizing is handy, and tends to be what AI is best at — but I also wanted to see if Claude could take some actions for me. The biggest issue for my inbox is the deluge of press releases, which can reach several hundred per day, and which all need organizing ready for reviewing.
While many of them aren't useful or relevant, some of them are, and it's applying this kind of discernment that I'm cautious about AI (or any kind of assistant) being able to manage. However, Claude proved adept at spotting which emails were press releases, and which were from companies or people I knew well.
Even better, it could apply my 'PR' tag to all the relevant emails that didn't already have it, ready for me to sort later. If you want, you can confirm each action manually, or have Claude work through them as a batch. Claude was able to do this quickly and accurately for me, which genuinely saves me time.
It even did a decent job at picking out the more worthwhile press releases from the general pile, and summarizing what was new in them. This isn't something I think I'd give over to AI entirely, but it's a handy way of quickly seeing if there's anything I missed, or getting an overview when I'm pushed for time.
I also liked the way Claude could pick out all the services I'd subscribed to recently — an occupational hazard for a tech journalist — and remind me to close them. As far as I could tell it didn't miss too much, and what I've seen so far has encouraged me to explore what else Claude might be able to do in my inbox.
There were occasional missteps, like when Claude told me I needed to reply to a review request email when I'd already sent a response to it, but they were few and far between. The usual 'check AI's working' applies here too, but it gets enough right that it ends up in credit as an inbox assistant that's actually worth keeping around.
