The cubicle was a building-block of the 20th-century economy. The agent is the building-block of the 21st. My thesis on what comes after — and why I'm building TrainYourAgent on the side of the humans, not against them.
I spent two years of my early career in a literal cubicle in a literal industrial-operations office. Forty hours a week processing routing tickets, fielding floor calls, escalating shipment delays, plugging numbers into the right cells of the right spreadsheets so that the people upstairs could make decisions. I was the substrate of the operation. The decisions weren't mine. The judgment wasn't mine. I was a node in a process that someone else had designed.
That kind of work — the substrate work — is what the AI agent is for. And the next 10 years of the economy are going to be defined by how gracefully or how clumsily we make that transition.
This is the thesis I'm building TrainYourAgent on.
A companion post, "How I Built a $20K MRR AI Agency in 12 Months (The Real Story)," used to sit next to this one and now redirects here. It was not the real story. The revenue figures in it were invented. There were never 14 active accounts, the month-by-month MRR ladder ($4,497 in September, $9,400 by Christmas, $14k in March, $20k in April) was fabricated, and so were the 78% gross margin, the $11k/mo retained earnings, the five partner agencies, the eleven partner-sourced deals, the four customers who took a booking-rate guarantee, and the eleven case studies. The pricing tiers it quoted ($799 / $1,499 / $2,500) were never our pricing either.
I deleted it rather than editing it quietly, because it was live long enough to be read and because a fabricated revenue claim is not the kind of thing you fix with a footnote. If you made a decision based on that post, I am sorry.
What is actually true and checkable about this business is on the about page and the results page, and the pricing is published in full on the pricing page. The thesis below is an argument, not a track record, and it should be read as one.
The cubicle, the call center, the open-plan office — these weren't just architectural choices. They were the physical instantiation of a particular theory of work: that complex coordination problems are best solved by densely packing many humans into the same building and routing standardized tasks through them.
It worked, for a while, because human labor was the cheapest available substrate for that kind of routing. A human routing ticket cost $0.50 to process. A human answering a phone cost $11 an hour. The economics of every white-collar business in 1990-2015 were premised on those unit costs.
The cubicle was the most efficient substrate we had. The work that filled it wasn't necessarily the work that was most rewarding for the human doing it — but it was the cheapest way to do it given the technology available.
In 2026, that same routing ticket costs $0.0006 to process via a Claude-backed agent. That phone call answered by Vapi + Eleven Labs + a tuned prompt costs $0.43 instead of $11.
The unit-cost gap is three to four orders of magnitude. And the gap is widening, not closing, because model prices keep dropping while wages keep rising.
You don't need to be a futurist to see what happens next. The substrate of white-collar work shifts from human to agent. Not all of it, not overnight, not without disruption — but the direction is set. The economics will not be overruled.
The mistake that doomers and boomers both make is treating this as a simple substitution: agent in, human out. That misreads the shape of the change.
Here's what actually happens when we install an agent at a typical TrainYourAgent customer:
Net staffing: -2 part-time roles, +0 senior roles (the senior was already on payroll, just promoted). Net cost to the business: -$57,000/yr. Net revenue uplift from better booking rate: +$120,000/yr.
The two part-time receptionists who got displaced — where do they go? The honest answer is: it depends. Some of them retrained into the senior front-desk role at the next dental office down the road, which is now hiring at the higher rate because the elevated role is what the new market is paying for. Some left for adjacent jobs. Some left the workforce. The transition wasn't free for them.
But the agent didn't "take their job." The agent took the parts of their job that were the lowest-leverage part of being a human. The parts that didn't require judgment, didn't require warmth, didn't require expertise. The parts that the cubicle was always a bad container for in the first place.
Look at what the typical receptionist job involved in 2015:
None of that is fulfilling work. None of it builds skill. None of it develops the human doing it. It is, in the precise sense, substrate work — work that exists because the system needs a body to route through, not because the body learns or grows from the routing.
What we're freeing humans from is not work. We're freeing humans from being substrate.
The other side of every install — the part the doomer narrative misses — is the work that the remaining humans actually get to do:
This is higher-skill, higher-paid, higher-meaning work. It's the work humans are uniquely good at. It's the work the cubicle suppressed because the operations couldn't afford to pay for it at scale.
The 21st-century white-collar role that emerges isn't "the receptionist replaced by AI." It's the front-desk specialist who runs five locations from one screen, supervises three agents, and gets paid like a junior manager.
That's not a worse job. It's not even a comparable job. It's a different one.
I'm building TrainYourAgent the way I am — services-led, retainer-driven, vertical-specialized, with a human always in the escalation loop — because I genuinely believe the right shape of this transition is the one where humans move up the value ladder, not the one where humans get cut to save margin.
You can build an AI agency the other way. You can sell "fire your customer-service team" as the headline. Some agencies do. Their installs work for 6 months and then degrade because nobody on the customer side has the skin in the game to maintain the agent or handle the escalations. The customers churn. The agency builds a worse business.
The agency that wins long-term is the one whose customers' humans get better when the agent ships. That's the install that sticks for 5 years. That's the customer that refers three more. That's the business that compounds.
I don't want to wave away the dislocation. Real people will lose real jobs in this transition. The economic gains will not be evenly distributed. The 2-3 year window where the displacement is happening but the new roles haven't fully emerged is going to be hard for a lot of households.
A few things I think we owe each other in that window:
I don't have a complete answer to any of these. I have opinions and I have one small business trying to model the right behavior. That's all.
I think the next decade plays out like this:
This is the macro thesis. The micro version is: I'm building one company in this transition. I want it to be a company whose customers' humans are better off after the install than before. If I do that 1,000 times, we'll have moved the needle a little bit in the right direction.
The thing I'd tell you: don't build "fire your team" software. The total-addressable-market on "displace the human" looks bigger but the customer LTV is shorter, the political backlash is real, and you're betting against the better version of the future.
Build "elevate your team" software. Build it for the SMB owner who wants their best receptionist to become a $90k/yr operations manager running three locations from one screen. Build it for the small law firm that wants its paralegal to do more matter-strategy work and less intake-form work. Build it for the customer-success team that wants to be customer-strategy advisors instead of ticket-clearers.
The agent installs that survive are the ones where the customer's humans love the change.
The thing I'd tell you: don't ship the agent until you've named the new role. What does the receptionist become after the agent ships? What does the support rep do with the time the agent freed up? What's the new title, the new compensation, the new responsibility?
If you don't have an answer, the install will fail politically inside 90 days even if the technology works.
If you do have an answer, the install becomes a force multiplier and your team thanks you for it.
I'm not building this company because I love agents. I'm building it because I want to see this transition go well — for SMBs, for their teams, for the customers they serve. The agent is the means. The humans-on-the-other-side becoming better at the parts of work that matter — that's the end.
If that's the kind of company you'd want to build with, or buy from, book a call. Even if you don't end up working with us, I'm happy to talk through where in this transition your specific business sits.