Last month I went to an event in San Francisco with two women I met in a comment section.

Kylee Lessard and Nisha Garg commented separately on something I’d shared on Substack . I replied to both, and both turned into private messages. Lessard mentioned that if I was ever in San Francisco she’d love to get together, and that if there happened to be an event happening while I was out there, she’d like to invite me as her guest. She’d founded Women + AI , a digital media and community platform that started with events as a way to build community trust. As luck would have it I already had a trip booked, and an event happened to be scheduled the week I was going to be there. Garg, it turned out, was the SF chapter co-lead and hosting it.

The story gets funnier, so stay with me.

A few days after the event I spoke with each of them on Zoom, separately, and the origin story got even better. Lessard and Garg had worked at the same company at the same time, in the same office no less, and had never met. It took until early 2026, at a small dinner Lessard organized, and only because a colleague pushed to get Garg through the door. And to deepen the intrigue, I was at that same company during that same stretch, just in a different office.

I would love to tell you that I engineered this romcom worthy setup, but I didn’t. I just started sharing on Substack, only a couple of weeks prior no less. It was that simple.

So there were three of us at one company with a very large pile of mutual connections between us, and it still took a comment section on social media, a colleague’s favor and a non-company event to put us all in a room together. Every part of that was relatively slow, unfolding over something like six months, and none of it could have been sped up.

Which is interesting, because with AI tools now at everyone’s fingertips, almost everything else about professional life seems to have gotten dramatically faster.

The Sorting Mechanism Broke

Nobody has ever had a Sorting Hat. Apologies to the Harry Potter fans out there still holding out hope of finding one at an estate sale.

There is no object, at least as far as I know, that miraculously reads what a person is made of and announces where they belong, instantly and incorruptibly. What the working world had instead was a much cruder proxy, effort.

Effort was information. It showed up as a researched pitch, a detailed cover letter, a cold message somebody mulled over for a week before sending, a conference application, or an unsolicited proposal. The outcome of those things was never actually about the writing but the effort it took to do them, to find somebody’s contact information, and then to have the audacity, if you will, to send it. All of it was proof that a person had spent something they couldn’t get back. Their time.

Now let me transport you to the late 1990s, when the most prized gift a person could receive was a mixtape. The value in it had nothing to do with the songs. It came from the fact that somebody sat in front of a boombox for two hours, hitting record, timing fades, and writing out the track list by hand, possibly in a gel glitter pen. The final product they handed over wasn’t a cassette. It was an afternoon of their life, their energy and their enthusiasm, sealed ever so preciously in a plastic case.

A playlist now takes just a couple of minutes to make and is better in every possible way, which is exactly why its value is less. You can build one and text it to somebody before your coffee gets cold. It’s still a lovely gesture, but not on the same level as the masterful mixtapes of the 1990s.

The professional version of that collapse is similarly measurable, and hiring is where you can see it most clearly. LinkedIn recorded job applications rising more than 45% year over year , running at 11,000 applications per minute as of June 2025, roughly 74% of U.S. job seekers now use AI to polish their applications, and nearly half say they are submitting more applications this year than last.

That volume is not a surge of ambition. What we’re seeing is the result of effort becoming cheaper, and in plenty of cases, free.

When the price of a signal drops to almost nothing, the signal stops carrying information. Every message is researched now, every follow-up arrives warm and on a sensible cadence, and inboxes are quite literally overflowing with emails that would have been genuinely impressive a decade ago. Every one of them is another one in a tidy, polished pile, and there is no way left to stand out inside it.

This isn’t a volume-of-bad situation. It’s that it all got better with better grammar and better timing and basically better everything, and somewhere in the middle of all that improvement it stopped sorting anyone. It stopped being the magical hat we’d all been relying on.

Which raises the question of what happens to a job market built almost entirely on that sorting.

We were led to believe that automated screening would finally read people correctly, an actual Sorting Hat for employment at last. But a tool that evaluates polish and how closely a resume matches a job description cannot tell anyone apart once that polish is free.

Daniel Chait, chief executive of the applicant tracking company Greenhouse, described where that leaves both sides: “Trust is at an all-time low for both job seekers and recruiters.” Roughly 65% of surveyed U.S. hiring managers said they had caught applicants using AI deceptively, and only 8% of surveyed job seekers said AI makes hiring fairer.

Kylee Lessard, the product marketer who founded Women + AI, has noticed the same thing from the recruiting side of the table. “AI can rapidly make any person look qualified for a job on paper,” she said. “And so, the sentiment, at least in the Bay Area, is that recruiters have had to go back to referrals, and hiring managers are really trying to pull people that they know and trust directly.”

That instinct has always had numbers behind it. Ashby, looking at more than 38 million applications across 93,000 jobs from 2021 through 2024, found that referrals account for roughly 1% of all applications and that 40% of referred candidates get from application to interview, against 3% of inbound applicants. Among those who reach an interview, 16% of referrals go on to an offer, against 6% of inbound. One percent of the pile, and more than 13 times the odds of a first conversation. That data ends before the AI surge, so the gap was there already. What’s new is the size of the pile it’s hiding in.

The behavior is showing up more broadly too. In Greenhouse’s survey, 68% of surveyed U.S. hiring managers said they are more involved in hiring than they were a year ago, and 39% are running more in-person interviews.

Hiring is the easiest place to see this, but it’s creeping in basically everywhere. The same collapse is running through every channel where people used to distinguish themselves by trying harder, whether that’s pitches, proposals, fellowship applications or cold introductions. What people used to call networking was one crude instrument for doing this, and it was never a particularly good one, but at least it ran on something that cost the sender.

Everything Got Faster Except One Thing

If you made a list of what AI has compressed, it would run longer than Santa’s, but unlike Santa’s it isn’t getting checked twice. Research that took a week now takes an afternoon, a first draft takes seconds, and design, analysis, translation, summarization, code and all the polish on top of it, which is basically the sprinkles, have gone from being the work to being a light touch.

And the part that arguably matters most is what happens at the very beginning, because starting is where most people lose. Not the doing, the starting, which is the blank page and the whole mental undertaking of a thing before a single word of it exists. I personally have problems with this. As an example, it took me more than a week after the San Francisco event to start writing this article. And it wasn’t because I wasn’t excited about it, I just knew it was going to be a big undertaking. If you have ever sat in front of an empty document fora chunk of time and then decided to reorganize a drawer instead, you can probably relate and already understand the research I’m about to cite. Piers Steel’s meta-analysis of procrastination research , published in Psychological Bulletin , found task aversiveness to be among the strongest predictors of whether a person delays at all. Making the first step less unpleasant takes that predictor off the table, and that is exactly what these tools do, for everyone at once.

Now make the second list, of what AI has not compressed, and it is dramatically shorter. A relationship built slowly cannot be rebuilt quickly, not by anything, at any price, by anyone, because there is no model, no budget and no strategy that produces a person who has known you for a period of time, possibly since before you were interesting.

Duration is not a task, it’s an accumulation, and it compounds over time. And that makes it one input with no shortcut, so it has effectively become the last place real effort is still visible.

That asymmetry is what I want to do the corporate double-click on. When everything else compresses and one thing does not, the incompressible thing becomes the remaining differentiator by default, not because it’s noble or because human connection is beautiful, though I’d argue it’s both, but because it’s the last thing left that cannot be generated.

What accumulates is not affection, although for some of us maybe it is. It’s the time a specific person has spent on you, usually in increments small enough that they don’t register as real effort. That’s the first expense, and I don’t think any of us are tracking how many times we send a podcast recommendation. A happy birthday message, a quick note on someone’s post, a quick call to catch up, a chat at an event when you see each other in passing.

But those small, seemingly insignificant things, are what make the second expense possible, because when that person thinks about recommending you, they are spending something else entirely. They’re putting their own judgment and their own reputation on the line, and if you turn out to be an epic disappointment they’re the one who looks like a poor judge of people to their boss, to their industry friends, to whomever. That kind of thing follows a person around a company like the smell of something spoiled in the back of your fridge that you can’t quite get out. It’s slow to build and impossible to get back.

It’s rare that anyone would spend that on a stranger. People spend it on somebody they know, not necessarily a friend but someone they’re at least familiar with, or whose work they have watched long enough to feel confident betting on, which is exactly what all that uncounted time was buying the whole way along.

Marketers have a pair of terms for the two halves of this. Mental availability is whether you come to mind when a need arises, and physical availability is whether anyone can actually get you once they do. A brand everyone loves and nobody can buy does not grow. It’s the cereal three separate people have told you to try that your grocery store has never once carried. So you stop looking for it. The same is true of a person.

So two separate things have to be true before anything reaches you, and most people are only working on one of them. You have to come to mind, and you have to be reachable by the people in the room where the decision gets made. You with me?

And this is exactly where AI has done its damage, because it flooded the first one and left the second one exactly where it was. Anyone can now put your name in front of anyone, at volume, for nothing, which means coming to mind through a machine-generated message has stopped meaning anything at all. What no tool can manufacture is somebody who is already standing where the decision happens and is willing to turn around and say your name.

Employment Used To Supply This For Free

The reason any of this feels new is that most people never had to build either kind of availability for themselves. Why not, you ask? Because an employer manufactures both as a byproduct of operations. It puts people together across functions they didn’t choose and wouldn’t have picked, over long stretches, and it shoves them together in virtual and in-person rooms where decisions get made for no better reason than that they’re on the same org chart.

Nobody experiences that as a benefit surprisingly. I surely never did. It pops onto your calendar in the form of meetings, which most people detest, including the recurring one you have been trying to decline since 2023. But depending on which meetings, it’s arguably building exposure that would take an unaffiliated person an enormous amount of time to manufacture on their own.

Take the job away, or the traditional W-2 version of it anyway, and the whole apparatus topples like Jenga underneath you. Which is why the panic that follows a layoff or a resignation is so specific and so disproportionate.

When I started putting myself out there, writing and sharing on social media and now on Substack, it stemmed from exactly that uneasiness. If the only people who can vouch for me are the people I work with day to day, what happens if I get swept up in a layoff?

What I was really asking, without having the words for it at the time, was whether I had any availability of my own. Plenty of people knew who I was, and almost all of them sat inside the same company structure, which meant that the moment I left it I’d come to mind for fewer and fewer people, none of them positioned anywhere new.

Everything compressible has been compressed, and the only remaining differentiator is something that accrues slowly and may not produce a result quickly, or at all, so the advantage goes to whoever can afford to spend the hours without knowing whether they’ll ever count. The input is no longer the product. The input is the waiting, and waiting is not something everyone has equal amounts of to give or patience to take it on.

While not the main point of this piece, there is an important point here that needs to be called out. A system that rewards accumulated time favors the people who have time to accumulate. People who were already inside the rooms, who were not working a second shift at home with caregiving responsibilities or a first job that ran late, who could afford to spend hours on something with no promised return. The old mechanism at least let a stranger buy their way in with effort, which made it a bad system but an open one. This one is harder to enter from the outside and much easier to keep once you are in. While I’m saying it is what’s left is not the same as saying it is fair.

Not so shockingly none of this is a satisfying system. Accumulated exposure is slow and unprovable, there’s no guarantee of a result, there’s no dashboard for it and no line for it in a weekly update, which for some of us is functionally the same as saying it never happened. From the outside, frequently from the inside too, it can feel like you’re running a marathon on a hamster wheel.

The countable version still exists though, which is a trap. You can still send the messages, still attend the events, still log the coffees and spit out an impressive number for anyone who asks. Those numbers just aren’t attached to anything meaningful anymore. Meanwhile the part that matters accrues somewhere you can’t see, on a schedule you don’t control, and it may not pay out for a very long time, or at all. I’d wager this is not at all satisfying for the data fiends out there.

What does all this mean? Well, it’s not enough for people to know your name. Tons of people can know your name and be lightyears away from anything that matters to you. In other words, it is entirely possible to be a household name inside one household. Far less impressive and beneficial.

And the inside keeps getting smaller. Gusto’s payroll data put the promotion rate at 10.3% in May 2025, down from 14.6% in May 2022 and below where it sat in 2019. For workers between 25 and 34, the group most often told to be patient, it’s 14.5%, down from 16% to 17% before the pandemic. Fewer rungs inside means more of whatever happens next depends on somebody off that ladder knowing you well enough to say so.

Nobody hands you the version of this that a job used to hand you. Which means the work now looks like nothing while you’re doing it, pays nothing up front, and is the only part of any of this entire system that still has to be done by a person. That last part might be moderately comforting.