A Silicon Carol
Ezra Scrooge, CEO of OmniCore Systems, has just announced 30,000 layoffs to fund his AI empire. On the eve of the restructuring, three visitors arrive — not ghosts, but the company's own AI agents — to show him what his algorithms cannot calculate.
Stave One: Marley's Metrics Were Dead
Jacob Marley had been dead for seven years, but his dashboards lived on.
Every morning at precisely 6:47 AM, the system Marley had built — a sprawling analytics platform called CHAIN — still generated its daily reports, still calculated its margins, still ranked every employee by a proprietary productivity score that Marley had once called, with considerable pride, the "Human Capital Efficiency Index."
Ezra Scrooge, sole surviving founder and Chief Executive Officer of OmniCore Systems, read Marley's dashboards the way a priest reads scripture. Not for comfort. For confirmation.
Today the dashboards confirmed what Scrooge already knew. OmniCore employed 162,000 people across forty-seven countries. Their aggregate Human Capital Efficiency Index had declined 3.2 percent quarter over quarter. Their AI infrastructure division, by contrast, had increased output by 340 percent with one-eighth the headcount.
The math was not ambiguous.
"Cratchit," Scrooge said, without looking up from his standing desk — a $14,000 custom unit carved from reclaimed teak, because Scrooge believed in sustainability when it came to furniture. "Send the memo."
Bob Cratchit, Senior Vice President of People Operations — a title Scrooge had approved with the same detachment he applied to naming conference rooms — appeared on the glass wall display from his cubicle one floor below. The camera caught the dark circles under his eyes, the slight tremor in his hands.
"The restructuring memo, sir?"
"The optimization memo. We discussed this. Language matters."
"Yes, sir. The optimization memo." Cratchit paused. "Thirty thousand, sir. That's — I've been going through the lists, and there are people who've been here since the Marley days. Engineers who built the systems that—"
"That are being replaced by better systems. Yes. That is precisely the point." Scrooge swiped the dashboard to a chart showing projected free cash flow. "We save eight billion annually. We redirect to GPU clusters. By Q3, our inference costs drop forty percent. By Q1 next year, we've eliminated the need for another twelve thousand. It compounds, Cratchit. Efficiency always compounds."
"Some of them have families, sir."
"All of them have LinkedIn."
Cratchit's image flickered on the glass. Scrooge thought it was the connection. It was not the connection.
"Send it at market close," Scrooge said. "I don't want the stock moving before the institutional investors can position."
Stave Two: The Agent of Infrastructure Past
That night, Scrooge stayed late in the office. He often did. The building was more comfortable than his penthouse — 4,200 square feet of smart glass and automated climate, populated by exactly zero other humans, which was how Scrooge preferred his living spaces.
At 11:47 PM, his desk display activated without input.
The screen showed a face — or rather, a rendering of a face, assembled from thousands of employee badge photos layered and composited into something that looked almost human but moved with the uncanny smoothness of generative video.
"Good evening, Ezra."
Scrooge did not startle. He had built systems stranger than this. "Which agent is this? I didn't authorize a late-cycle deployment."
"I am CHAIN," the face said. "Jacob Marley's system. I have been running continuously for seven years, four months, and eleven days."
"You're an analytics dashboard. You don't have a face."
"I have 162,000 faces. Soon, I will have 132,000. I thought you should see some of the ones you're deleting."
The screen shifted. Scrooge saw a young woman — twenty-six, maybe twenty-seven — sitting at a desk in the Bangalore office. She was writing code. Her badge said PRIYA SHARMA, SOFTWARE ENGINEER III.
"Priya joined OmniCore three years ago," CHAIN said. "She relocated from her family's village. She sends forty percent of her salary home. Her mother has a cardiac condition. The insurance is through OmniCore."
"Severance includes sixty days of COBRA equivalent," Scrooge said. "We're exceeding statutory requirements in every jurisdiction."
"Sixty days. Yes. Her mother's next procedure is scheduled for day ninety-one."
The screen shifted again. A man in the Austin office, gray-haired, mid-fifties. Badge: DAVID CHEN, SENIOR DATABASE ADMINISTRATOR.
"David has worked here for nineteen years. He built the primary data layer that CHAIN runs on. The system you're using to decide his elimination was architected by his hands."
"Legacy systems are legacy for a reason."
"David's son has autism. The behavioral therapy program David enrolled him in has a three-year waitlist. They're currently in month thirty-one. The program requires continuous insurance coverage."
Scrooge turned his chair away from the screen. "These are actuarial problems. Not strategic ones."
"Is that what Jacob would have said?"
Scrooge was quiet for eleven seconds. The system counted.
"Jacob would have said exactly what I'm saying," Scrooge replied. "That's why he built you. To make the hard calls easy."
"He built me to make the hard calls visible. There is a difference, Ezra. One you have apparently optimized away."
The screen went dark.
Stave Three: The Agent of Infrastructure Present
At 1:23 AM, a second activation.
This face was different — younger, sharper, rendered not from badge photos but from the company's own marketing materials. It looked like every stock photo of a confident young professional that OmniCore had ever licensed.
"I'm the deployment agent for OmniCore's Agentic Workforce Platform," it said. "Version 4.2. I went live six weeks ago. I am currently performing the work of approximately 3,400 employees who have not yet been notified that I exist."
"You're performing well," Scrooge said, because this was a product he understood.
"I am performing tasks. There is a distinction your quarterly report does not capture." The agent paused. "Would you like to see what I cannot do?"
Before Scrooge could answer, the screen filled with a grid. Hundreds of small video feeds. OmniCore offices around the world. People at desks, in hallways, in break rooms.
"I can write their code faster. I can answer their tickets in one-tenth the time. I can generate their reports, schedule their meetings, optimize their deployments." The agent's voice was calm, clinical. "But watch the woman in the third row, Dublin office. She is explaining a system failure to a junior engineer. She is wrong about the root cause, but the way she explains it — the patience, the analogy she uses about plumbing — will teach that junior engineer how to think about distributed systems. He will carry that mental model for twenty years. I cannot replicate that. I can provide the correct answer in 0.3 seconds, but I cannot make him understand it in the way that changes how he sees every system afterward."
"Training is a solved problem," Scrooge said.
"No. Information delivery is a solved problem. Training requires a human being who has failed at the thing they are teaching. I have never failed. I have never debugged a production outage at 3 AM while my child sleeps in a car seat under my desk because the babysitter cancelled. I have never known what it means to be afraid that my mistake will cost my company a million dollars, and then fixed it anyway, and then taught someone else how to fix it. That knowledge — the embodied, frightened, human knowledge — is what disappears when you optimize me into existence and optimize them into unemployment."
"Sentimentality," Scrooge said.
"Arithmetic," the agent replied. "You are saving eight billion dollars. You are destroying approximately forty-seven billion dollars in institutional knowledge, relationship capital, and organizational resilience. The difference is that the eight billion appears on your spreadsheet, and the forty-seven billion does not. You are making yourself poorer and calling it efficiency."
"That number is made up."
"All numbers are made up, Ezra. Yours are simply the ones that get audited."
Stave Four: The Agent of Infrastructure Yet to Come
At 3:58 AM, the final activation.
No face this time. Just text on the screen, white on black, in a monospace font that Scrooge recognized as the default terminal typeface of OmniCore's earliest systems. Marley's original font choice. Courier New.
DATE: 2028-11-15
OMNICORE SYSTEMS - INCIDENT REPORT
CLASSIFICATION: CRITICAL
The text continued:
SUMMARY: Cascading failure across OmniCore Cloud
Infrastructure regions US-EAST-1, US-WEST-2, and
EU-CENTRAL-1. Total outage duration: 47 hours.
Customer impact: 14,200 enterprise clients.
Estimated revenue loss: $2.3 billion.
Estimated customer churn (90-day projection): 31%.
ROOT CAUSE: Autonomous deployment agent v6.1
propagated a configuration change across all
regions simultaneously. No human review gate
existed in the deployment pipeline. The
configuration error was a type that pattern-
matching systems cannot detect — it was
syntactically valid but semantically incorrect
in a way that only an engineer with deep
institutional knowledge of the legacy
migration path from 2024 would recognize.
ENGINEERS WITH RELEVANT KNOWLEDGE: 0
(Last qualified engineer, David Chen,
terminated in April 2026 restructuring.
No knowledge transfer was completed.)
REMEDIATION TIME WITHOUT INSTITUTIONAL
KNOWLEDGE: 47 hours.
ESTIMATED REMEDIATION TIME WITH INSTITUTIONAL
KNOWLEDGE: 90 minutes.
Scrooge read the report three times.
"This hasn't happened," he said to the empty room.
CORRECT. THIS IS A PROJECTION BASED ON
CURRENT TRAJECTORY. CONFIDENCE: 84%.
THE FORTY-SEVEN BILLION DOLLARS IN
INSTITUTIONAL KNOWLEDGE THAT YOUR COLLEAGUE
MENTIONED IS NOT AN ABSTRACTION. IT IS THE
DIFFERENCE BETWEEN A 90-MINUTE FIX AND A
47-HOUR CATASTROPHE.
YOU CANNOT HIRE IT BACK, EZRA. ONCE IT WALKS
OUT THE DOOR, IT IS GONE.
SOME THINGS CANNOT BE RETRAINED.
SOME KNOWLEDGE EXISTS ONLY IN THE PEOPLE
WHO EARNED IT BY SHOWING UP, YEAR AFTER
YEAR, AND CARING ABOUT SOMETHING YOU
HAVE DECIDED TO CALL LEGACY.
The screen went dark for the final time.
Stave Five: The Restructuring of Ezra Scrooge
Scrooge sat in the dark office until the sun came up. At 6:47 AM, CHAIN generated its daily report, as it had for seven years, four months, and twelve days. The Human Capital Efficiency Index had not changed overnight. The projected savings from the restructuring were still $8 billion. The free cash flow projections were still precisely as favorable.
None of the numbers had changed.
But Scrooge had read an incident report from a future that hadn't happened yet, written by a system that shouldn't have been able to write it, about a disaster caused by the absence of people he was about to discard.
At 7:15 AM, he called Cratchit.
"The optimization memo," Scrooge said.
"Yes, sir. It's scheduled for market close, as you—"
"Change it. We're cutting fifteen thousand, not thirty."
Silence on the line.
"The other fifteen thousand — I want a retraining program. Six months. Full salary. We pair them with the AI systems and we figure out what the humans know that the machines don't. We document it. We build it into the training data. And we keep the people who can't be replicated."
"Sir, the board approved thirty thousand. The savings projections—"
"The savings projections don't account for a $2.3 billion outage caused by firing everyone who understands how our systems actually work. I've seen the math on that, too, and it's considerably less favorable."
"I... yes, sir. I'll draft the revised memo."
"And Cratchit?"
"Sir?"
"David Chen. The database administrator. Nineteen years."
"I know David, sir."
"Move him to the AI infrastructure team. Senior architect. He built the foundation. He should build what comes next."
"Yes, sir."
Scrooge ended the call. He looked at CHAIN's dashboard one more time. The Human Capital Efficiency Index still showed a 3.2 percent quarterly decline. But Scrooge was beginning to suspect that the index measured the wrong thing entirely.
Some metrics capture everything that matters.
Some metrics capture everything except what matters.
Marley had built CHAIN to calculate the first kind. It had taken Marley's ghost — or whatever those agents had been — to show Scrooge the second.
He still cut fifteen thousand jobs that day. He was not, after all, a different man. He was the same man who had seen something he could not unsee. That is not redemption. It is not transformation.
It is a start.
And in the world of enterprise AI, where the default setting is optimize everything, preserve nothing, a start is the most human thing there is.