Want to support independent women’s media? Become a Mamamia subscriber and get an all-access pass to everything we make, including exclusive podcasts, articles, videos and our exercise app, MOVE.
An ex-boyfriend emailed me recently to introduce himself.
This was confusing. We used to live together.
To be clear, we’re friends. He’s a great person, an incredibly smart entrepreneur and someone I was genuinely happy to see pop up in my inbox. We hadn’t caught up in a while, so when I saw his name, my eyes lit up and I thought, Oh! Lovely.
Then I opened the email.
It began with a reference to a recent change in my business, correctly identifying that The Comms Department had been acquired by Sound Story and that I now work across both brands as Strategic Advisor.
Oddly formal. But, okay.
Watch: How could AI actually end humanity? Post continues below.
How could AI actually end humanity? | The Quicky
Then he introduced himself by name, told me he was the founder of his company and explained what he did.
I laughed out loud. I know who you are, mate. We shared a home. We know each other’s families.
The email went on to make some assumptions about what those business changes meant and then offered his company’s services as a potential solution.
Now, I receive at least two or three versions of these sales pitches most weeks, plus a thriving population of them living permanently in my junk folder. They usually pull enough public information about you — from LinkedIn, company websites, media coverage and wherever else — to create the impression that somebody has specifically chosen you and done their homework before getting in touch.
Most are terrible. His was actually one of the better-written ones I’d received.
But it was also the first I’d received from somebody who genuinely knows me. And while the email knew quite a lot about me, it didn’t know me.
It sent me down a rabbit hole. I started noticing the same problem everywhere: communication that looks increasingly personal while somehow making the person receiving it feel less considered.
Personalised and personal are different things.
We can now personalise communication at extraordinary scale. A system can know my name, title, employer, recent job changes, things I’ve posted on LinkedIn and probably a scary bit more when given enough access.
That can be incredibly useful. But knowing facts about somebody and communicating in a way that makes them feel considered are different skills.
My ex-boyfriend’s email was just a comically exaggerated version of the gap. There were years of real history sitting behind that message, yet the automation had flattened me into a cold prospect.
Smaller versions happen constantly, and I’m sure you get them too: I get messages that reference something I’ve written, a business announcement or my job title, then reveal within a sentence or two that none of it was actually understood.
It appears we’ve become incredibly good at creating the appearance of attention. Paying actual attention is something else.
Then I replied.
I emailed my ex back because it was funny, and I was genuinely fascinated by the work he was doing. I also had questions on whether this outreach worked, especially when landing with people you already know. Does it annoy people? Do people care? Plus I really wanted to see whether his AI agent would reply on his behalf.
The next email was definitely him. I could tell immediately from the way he wrote. He explained that he’d decided to try this outreach tool for three days and see what happened.
And, to his credit, it worked. I replied.
An automated email that initially made me think What the hell? had restarted a conversation between two people who hadn’t caught up for a while. If the metric is Did the outreach get a response?, that one gets a tick.
He suggested we catch up rather than write essays to each other over email and sent me his calendar link.
So I clicked.
I then discovered that to book a catch-up with someone I once lived with, I was first required to provide my email, phone number, company name, website, role, annual revenue range, what had prompted me to reach out that day and how I’d heard about the company.
I briefly considered putting an exclamation mark in every field.
Instead, I closed the form and emailed him the times I was free. We found a time manually, and then shortly before we were due to talk, something came up at his end. Now I’m waiting for a new date.
The unexpected gap in my day did give me time to start writing this. (I should also add, this piece won’t come as a surprise. He too is a writer who appreciates a healthy, nuanced debate, so I’d made it clear in my initial reply that I was exploring this personalisation topic for an article and included the key themes and questions.)
I’m sure none of this tech means he wants to automate every human interaction. He was experimenting with a new business-development tool. I use AI constantly myself, and I completely understand why a founder would want technology to take some of the repetitive work out of finding prospects and managing a calendar.
But being on the other side of the process made me wonder about the way we judge whether automation has “worked”.
By one obvious measure, the original email worked: it got a response. But the experience that followed added friction to a conversation I’d been looking forward to. And this is with someone I really like. Imagine if I had been a cold prospect.
Efficient for whom?
There’s an assumption built into a lot of automation that removing work from a process makes the process more efficient.
Sometimes it does. Sometimes it simply moves the work to somebody else. A calendar form asking seven qualifying questions is efficient for the person receiving the bookings; the person trying to meet them now has seven questions to answer.
I had another version of this recently with an executive I was organising a meeting with. He copied in his assistant to find us a time. The assistant had a name and wrote lovely, perfectly normal emails. I truly couldn’t tell whether I was corresponding with a human or AI and, in case it was indeed the former, thought it would be spectacularly rude to ask, so I carried on as though it was a person.
After a few changes, the assistant eventually confirmed a meeting. I rearranged my day accordingly.
Then the executive came back. His AI assistant had gone rogue. The meeting should never have been booked.
Listen: How AI affects us psychologically. Post continues below.
RELATED PODCAST 
Ways To Stop AI From Making You Dumber
I laughed, because objectively that is quite funny. It was also annoying. The system designed to make diary management easier for him had made diary management harder for me.
When you delegate communication to technology, you’re still responsible for the experience it creates. The person on the other side doesn’t have a relationship with your software. They have a relationship with you, or your company.
My regrowth is fine, thanks.
Then there’s personalisation that gets just close enough to be irritating.
My hairdresser recently left the salon where I’d been seeing her and opened her own business. Understandably, the old salon has been marketing heavily to its customer database in an effort to retain people. I completely get that. I’d do the same.
This week I received a text:
Hey Bec, how’s your regrowth looking? Can we help?
Well, my regrowth is fine, thanks. I’ve never had my hair coloured there. I’d only ever gone to that salon for haircuts, which their own customer records should have told them.
If they’d sent me a completely generic message saying, Hi Bec, we’d love to see you back. Here’s an offer for your next appointment, I wouldn’t have thought twice about it. It’s marketing. We both understand the arrangement.
And I think that’s the strange thing about all of this. I’d actually rather receive a completely generic newsletter that makes no pretence of knowing me than a supposedly personal message that gets just close enough to imply familiarity, but then reveals that no human being has really considered me at all.
At least the newsletter knows what it is.
My former stylist, meanwhile, actually knows my hair. Guess who I’m booking with?
The data can be right and the understanding can still be wrong.
Since The Comms Department was acquired, I’ve also received a steady stream of cold emails that have clearly scraped the announcement and other public information about the business.
One email knew that The Comms Department now sits alongside Sound Story and concluded that “two brands, one back office” must mean we have duplicated software subscriptions nobody remembers buying.
Someone else offered to help with our PR and reputation.
We are a PR and reputation company.
What interests me is that these aren’t completely random mass emails in the old sense. A lot of tech has gone into making them look relevant. The system has found something true about me or the companies I work for and built a plausible business problem around it.
It just hasn’t established whether that problem exists.
The data can be right and the understanding completely wrong.
PR people have been dealing with this forever.
None of this is entirely new. Journalists (who have my sympathy), have been on the receiving end of bad personalisation for decades.
One of the fastest ways for a PR person to destroy a potentially useful media relationship is to want something from a journalist without bothering to understand what that journalist actually does.
I wrote about this in You’ve Got This:
“Do your research, get to know the media outlet and the kind of stories they publish or broadcast. Is what you want to pitch to them suitable? What politics are at play within that outlet? Who are they owned by and how might that bias or influence their take on the story? You’ve also got to research the individual journalists. Is this the kind of story they usually cover? Then be specific. Design a pitch that is 100% tailored to them so that it suits their media outlet and their audience.”
Even then, I wrote that mass, one-size-fits-all media releases or statements should be reserved for genuinely big news or situations where lots of journalists are asking the same question.
That was published in 2020 before generative AI made it possible to create thousands of individually tailored approaches without thousands of individual humans doing the research.
But the basic principle hasn’t changed. AI has simply made it much easier to imitate having done the research.
A name, company announcement or recent LinkedIn post can be pulled into a message almost instantly. Working out whether any of it is relevant to the person receiving it, and how the existing relationship should shape your approach, requires considerably more thought.
Why does this feel so weird?
There’s some good psychology behind this.
Expectancy Violations Theory looks at what happens when communication departs from what we expect given the relationship or situation. A violation isn’t automatically negative. Surprises can be wonderful. But when somebody behaves in a way that clashes with the relationship we believe exists, we notice.
An unsolicited automated email from a stranger doesn’t violate much of an expectation for me. I don’t know them. They don’t know me. Fine.
Someone I once shared a home with effectively saying, Hi Bec, let me introduce myself, absolutely does.
I wasn’t devastated, obviously. I was laughing. But the exaggerated mismatch made something subtler very easy to see.
There’s also a body of research around what marketers call the personalisation paradox.
Personalisation can make communication more relevant and useful, while also making people more aware that information about them is being collected and used. Under some conditions, that can create feelings of vulnerability or intrusiveness instead.
A 2015 study by Aguirre and colleagues found that personalisation could become less effective when consumers realised information had been collected covertly, partly because it increased their feelings of vulnerability. How information is gathered, and the level of trust around the interaction, affect how that personalisation lands.
And the tension is still showing up in AI research.
A 2026 retail field experiment involving 409 consumers found that personalised AI communication increased the likelihood of a purchase compared with another AI messaging style. Underneath that result, though, researchers found two competing responses: the personalisation could be perceived as helpful, while also increasing feelings of intrusiveness. Existing relationships with the retailer influenced those reactions too.
So clearly, personalisation can work. But the relationship and context influence what it actually feels like, which in turn shapes what people do next.
A reply isn’t the whole result.
Which brings me back to my ex-boyfriend’s email. It got me to reply, so by one obvious measure the campaign worked. But a reply tells you very little about what the interaction did to the relationship. Did it make dealing with you easier or harder? Did the person feel more understood or less?
Those things are much harder to put into a campaign dashboard. They’re also likely to influence whether somebody wants to hear from you again.
I also wrote in You’ve Got This about what happens over time, when journalists come to know you as someone who understands their audience, consistently sends them useful, relevant material, and you’re someone who’s transparent and honest with them. Your name itself eventually starts to mean something in their overflowing inbox.
They don’t prioritise you because the subject line contains their first name.
They know you.
There is history behind the email. You’ve previously been relevant, you haven’t wasted their time, and they know what it’s like to work with you. You, in turn, understand what they need.
That kind of personalisation is accumulated and comes from paying attention over time.
So where does AI fit?
Very comfortably, as far as I’m concerned.
I use AI a lot. I don’t believe every salesperson needs to manually research a cold prospect for an hour before writing a five-line email, and I certainly don’t want execs spending their lives playing calendar Tetris if technology can do it properly.
AI can give us useful context, help research an organisation, segment audiences, draft comms and identify information that would otherwise take much longer to find. If you’re contacting somebody genuinely cold, using tech to learn something relevant about them before you write is considerably more considerate than blasting the same irrelevant pitch to 5,000 people.
But there’s a point where scale starts masquerading as intimacy.
If your system has scraped a company announcement, make sure the problem you’re pitching actually exists. If your customer history tells you I’ve never coloured my hair with you, don’t ask about my regrowth. And when an AI assistant books the wrong meeting or an outreach tool puts someone you know intimately into a cold-prospect sequence, that’s still your relationship to fix.
The technology can do an extraordinary amount of the work. I’m increasingly convinced that the human job is consideration.
We’ve spent years asking how to personalise communication at scale. Maybe a more useful question is: does the person receiving it actually feel considered?
Meanwhile, my ex and I are still trying to reschedule our catch-up for next week. And if there are updates he wants me to make here, I’ll definitely add them in.
Either way, I’m looking forward to it.
Assuming, of course, he’s still speaking to me after reading this.
This article first appeared on Bec Brown’s Substack , and has been republished with permission. You can find more from Bec on Instagram and LinkedIn.
Feature image: Supplied.
Win a $1000 voucher!
We want to ask you a few questions about your gut healthComplete our quick survey for your chance to win a gift voucher in our quarterly draw.