I Tried Automating My Relationships Utilizing AI Agents. Muse and Dots Overstepped – CNET

Most folks enjoy milestones. Turning the yearly web page on birthdays, holidays and anniversaries is a chance for large events, exchanging good items and celebrating those we love. It may also be a minefield of missed connections and harm emotions in the event you neglect the dates.

If you’re ever despatched a “happy belated birthday” message, uncared for to recall your individual anniversary or just wrestle to be in contact with folks, this information may be for you. Granted, you’ll must put your belief in an AI agent to do the heavy lifting, however you would possibly by no means want to recollect to ship a card once more.

While AI might not perceive human connection and happiness, what it does nicely with is rote duties. I attempted to make use of ChatGPT and some different AI instruments to automate the maintenance of my private {and professional} relationships — however right here’s the place all of it went mistaken.

(Disclosure: Ziff Davis, CNET’s mum or dad firm, in 2025 filed a lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI methods.)

‘Happy belated birthday,’ no extra?

The penalties for being horrible at remembering birthdays improve as one will get older. Soon it’s not simply your good friend’s birthdays that slip your thoughts, but additionally their youngsters’ birthdays and your grandkids’ birthdays and on and on because the wheel of copy turns. 

I used ChatGPT — which seemed to be the one AI device among the many three I examined that would automate this whole course of for me — to establish key birthdays of family and friends and develop an automation system to verify I by no means miss one other vital date. 

Claude was in a position to entry and draft emails in addition to create birthday playing cards and greetings, however not automate the sending.

Gemini appeared prefer it may solely work with me within the chat to draft emails and generate customized playing cards, not really work inside my inbox. Only some folks have entry to Google’s AI agent, Gemini Spark, proper now.

Gemini/Screenshot by CNET

I licensed ChatGPT and its agentic providing, Dots, to look by means of my Google Calendar and Gmail to search out and match dates to the birthdays of the folks most vital to me. You can do that by navigating to your profile icon in ChatGPT, clicking Settings after which opening Plugins.

ChatGPT/Screenshot by CNET

I additionally gave ChatGPT authorization to learn, compose and ship emails on my behalf. 

Then I requested ChatGPT to look by means of my calendar and Gmail and discover as many vital birthdays because it may.

ChatGPT was in a position to make use of my inbox to boil down my mates’ and household’s pursuits to a broad, but particular, sufficient set of information factors to create customized digital birthday playing cards that really match all their personalities. Even the one good friend ChatGPT know a lot about nonetheless acquired a digital card that match her just about to a T.

ChatGPT/Screenshot by CNET

Automating all my love

Now comes probably the most essential aspect — establishing ChatGPT to robotically ship digital birthday playing cards and greetings to all of the vital folks in my life so I don’t screw it up. I examined this characteristic by asking ChatGPT to generate a digital card and message for my canine’s birthday and ship the e-mail to my husband.

ChatGPT/Screenshot by CNET

The chat restrict was reached right here, and ChatGPT additionally surfaced an alert that requested for my permission to ship an e-mail on my behalf. And whereas it may be good normally for an AI agent to ask to your permission earlier than sending an e-mail, I had tasked it with sending these particular emails by itself robotically every time.

Beyond my e-mail job, I additionally requested Meta Muse, the social media large’s AI agent, to ship memes and check-ins to a couple of my mates on Instagram. This ended up being extra hassle than it was value; the content material Muse chosen to ship alongside was based mostly on seemingly nothing particular, simply random posts sourced from across the platform that may be construed as mildly amusing. 

The backside line on AI brokers overstepping

This is the place issues get slightly fuzzy. The quantity of labor it took to optimize Dots and Muse so they might ship related, plausibly practical correspondence was not value it. If it takes extra time to overview, revise and approve a easy e-mail or DM than it will for me to do it myself, the entire prospect and promise of agentic AI falls flat.  

Ultimately, you must also bear in mind that when you’re working with an AI agent, it might additionally begin to overstep within the chat. I needed to inform each Muse and ChatGPT’s Dots to again off — as a result of whereas I’d given them permission to peruse my emails and social media accounts, I had solely requested them to finish sure, very particular, actions whereas doing it.

ChatGPT/Meta/Screenshot by CNET

Dot took it upon itself, completely unprompted, to alert me to a timing snag in my night, going by means of my emails and letting me know a film I’d purchased tickets for conflicted with my Japanese class that night time. This might need been helpful, had I not already despatched an e-mail to my instructor proper after shopping for the tickets, letting them know I’d be lacking class.

Muse began sending me unsolicited updates concerning the climate and a fireplace in Malibu. Both undertook relentless efforts to get me to make use of an increasing number of of their options whereas I had the apps energetic. 

Overall, the expertise didn’t go away me feeling extra related to something however the AI firms.

Rachel Kane

Contributor and former Senior Editor

Rachel is a freelancer based mostly in Echo Park, Los Angeles and has been writing and producing content material for practically twenty years on topics starting from tech to trend, well being and way of life to leisure and schooling. She’s at present a Professor of Practice at Arizona State University’s Walter Cronkite School of Journalism and Mass Communication, serving to to mould the brand new minds who will inherit the media panorama. She’s hoping to stop the singularity by being well mannered to chatbots and spends manner an excessive amount of time refining Midjourney prompts.

See full bio