I was there during the dramatic reshaping of the knowledge worker economy during the collapse of the dot-com bubble during the late 90s. I was relatively new to the industry, early enough in my career that most of what I understood about how work was supposed to feel was something I was still learning from the people around me.
We got our work done, but we would also joke around with each other, go to lunch together, hit mid-week happy hours with clockwork regularity, and bounce half-formed ideas off of each other. There were office romances and company rituals. We shared a common cause, and the people who had been there long enough to know how the place actually worked were our guides as we navigated the company culture.
Then one day, one of the people I worked with wasn’t there. No announcement, just a brief meeting, and the person’s desk was empty.
And they weren’t the last. As the fallout from the bursting of the dot-com bubble expanded, coworkers that I had bonded with would be there on Friday, and gone by Monday. I’d walk past a cubicle that used to belong to someone I knew and eventually stopped expecting to see anyone there. After a while, I stopped being surprised by it; that was its own kind of loss.

Years later, I’m watching the cycle repeat itself as AI reshapes the workforce (or serves as a convenient excuse for adjusting headcount because of previous over-hiring). The industry has changed, the mechanism of disruption is different, and the shape of it is the same: a period of rapid change, an organizational response that reaches for efficiency, and a workforce left to figure out what was actually lost in the process.
The difference this time is that the technology driving the disruption is also inside the building. AI tools are being positioned as the solution to the headcount reductions they helped justify. Do more with less, because now you have the tools to make that possible, right? What that argument skips is unquantifiable culture memory that was in the headcount, something intangible that a tool can’t replace.
Most organizations describe culture as a set of values, typically a beautifully-crafted statement on the About page. That description captures a very real thing that could be described as a “way of being” for an organization, but I think it misses the mechanism.
Culture is less a collection of values than a kind of permission system. It tells people, without anyone having to say it directly, what is actually allowed here: to not know something yet, or to be uncertain about a direction, and struggle to say so. Remember the people who had been there long enough to know how the place actually worked? I meant that both in terms of process and, more importantly, culture.
We shared a common cause, and the people who had been there long enough to know how the place actually worked were our guides as we navigated the company culture.
For example, everybody now knows someone they worked with, who they genuinely liked, who has been laid off (probably pretty recently, too). That carries with it a kind of grief that is felt both on the personal level as well as the organizational level. Think about this: organizations do not give explicit permission to grieve a layoff. They used to give it implicitly, through the manager who checked in, the or the coworker who created space for the conversation. The true culture that acknowledged loss as a real thing that happened to real people. When those carriers are gone, grief has no place to land. It goes underground. It resurfaces later as disengagement, or as attrition. It shows up as the particular flatness of a team that is performing without investing.
Organizations do not give explicit permission to grieve a layoff.
There is also something specific about the position of the people who stayed. There is the expectation, often unspoken, that they understand they were the fortunate ones and should act accordingly. That expectation forecloses the honest experience, which is a complicated mix of relief, guilt, and the ongoing work of doing a job that used to require more people than it does now. Asking someone to operate at a higher level while carrying all of that, without acknowledgment, is asking a lot. But that’s the new norm.
Cultural permission travels through relationships. We’ve all had coworkers who noticed we were just kind of off before we said anything. We’ve had managers who thankfully pushed back when leadership made an unreasonable ask. And there was always the person who had been there long enough to know which battles mattered and would straight-up tell you so. Documents and all-hands announcements just can’t do that. Cultural permission is carried by people, passed between them through time working alongside each other, and it’s ambient enough that you rarely notice it until it is gone.
Culture also moves through contact. Think of that ambient permission made visible through repeated behavior over time. When the surface area of contact shrinks because there are fewer people, and quite frankly, less slack (the lowercase “s” variety) in the day, there are fewer jokes, fewer lunches and mid-week happy hours, and definitely less half-formed ideas to bounce around. The transmission weakens. New employees get less of the true cultural undercurrents, and instead, they are handed documentation. That’s not the same thing; documentation doesn’t transmit permission, just information.
There’s also the permission to just…stop. I always remembered the colleagues who said it could wait until Monday (thank you for that!), or the manager who left at a reasonable hour, modeling permission behavior that it was actually okay to have a life. Without people to transmit these unspoken cultural norms, and AI tools are always available, and individual output is measurable in real time, the permission to stop becomes something each person has to manufacture alone. Most people are not good at that, especially under pressure. So they don’t stop. It’s much easier to borrow that permission from the culture around you, to look up and see that other people are leaving and take that as a signal that it is okay to leave. When the culture that carried that signal is no longer intact, people stay later than they should, work on weekends more than they mean to, and describe their exhaustion in terms of personal failure rather than structural reality.
[People] describe their exhaustion in terms of personal failure rather than structural reality.
Organizations can rebuild culture after it has eroded. They can be pretty intentional about it, make room for it, hire people who understand its value. What they can’t do is send a memo that restores that unspoken permission, or install a tool that tells people it is okay to not know something, to be struggling, to grieve a colleague who is gone. That transmission requires honest-to-goodness relationships. Relationships requires time and contact and a degree of inefficiency (a dirty word in 2026) that hyper-optimization is specifically designed to eliminate.
What gets optimized away, past a certain point, is more than a set of values. It is also the infrastructure through which people learn what they are allowed to be at work. And the people still in the building, doing more than they were hired to do, running workflows that were supposed to make the job easier, carrying losses the organization never acknowledged, are the ones rebuilding that infrastructure from scratch. Without being asked. Without being thanked for it. Often without knowing that is what they are doing.
That was true when the dot-com bubble burst at the start of my career. It is true now, in different orgs, with different tools, and the same cost to the people inside them.

The term “Morni” translates to “peahen” in Punjabi. In this propulsive track, Diljit uses this imagery to describe the elegance and charm of his love interest.
One of my all-time favorite bhangra songs is “Mundian To Bach Ke” (“Beware The Boys”), a breakthrough hit for Punjabi MC that was so popular, it even has a remix version with none other than Jay-Z.
I now have a new Punjabi obsession: “Morni” is a killer groove by the charismatic Diljit Dosanjh, a successful actor/musician who has been around for years, and now seems (rightfully) poised on the precipice of a Bad Bunny-level global breakthrough.
“My managers keep taking my copy that’s with them for final approval, running it through AI, and going ‘here, use this.’ A huge waste of time and effort on everyone’s part, not to mention demoralizing as hell.”
The central idea: when the person who is supposedly in a role because of their experience and intelligence is constantly outsourcing their instincts to AI, it says to a team, “I don’t trust my own opinions on humor, storytelling, and marketing”. Which, if that’s the case, wouldn’t that make those leadership jobs the most susceptible to be replaced by AI?
A Skill is a set of instructions - packaged as a simple folder - that teaches Claude how to handle specific tasks or workflows.
When asked about my AI workflow, the first thing I always recommend is to get comfortable using Skills to customize Claude (my LLM of choice).
Once I really understood the power of designing, refining, and maintaining agent Skills, I stopped trying to fit AI into my existing linear workflow, and started working in a completely new “hub and spoke” model. Think of writing Skills like you would write an incredibly thorough creative brief. The engineering group at Perplexity put it best: “If you write a Skill like you do code, you will fail.”
Thats it for now, thanks for stopping by 📬
And as always, thanks for being a local, and be sure to visit Rural & Co. to continue the conversation. You'll be glad you did.


