A large organization announced Friday that it had eliminated its entire two hundred dollars per engineer weekly AI inference expense. This cut AI spend to zero while keeping software delivery perfectly flat. Leadership attributed the achievement to a handful of simple tips that can be adopted by any company with salaried employees and a sufficiently selective dashboard.
The policy follows a four week internal study involving ten engineers. The study found that engineers using AI completed thirty percent more accepted work without increasing defects or review time. They did, however, consume two hundred dollars per person in inference, or two thousand dollars across the team.
Executives described the result as mixed.
A company spokesperson said, "Obviously, we were excited by the productivity. But then Finance asked why the engineers were leaving after only eight hours. Once we saw the unused inventory sitting between eight o'clock p.m. and six o'clock a.m., the solution became obvious."
Leadership agreed to share its seven step playbook. None of the tips requires new technology. Most require less.
Here is the first tip: turn off the tool before you finish reading the study. The company turned off the AI on Monday. Engineers were instructed to begin work at six o'clock a.m. and finish at eight o'clock p.m. Weekends were reclassified as "meeting free production zones." Lunch remains available as an asynchronous activity and will be delivered directly to each desk so leaving the keyboard does not interrupt flow. The cost of each meal is deducted from payroll, thanks to HR's quick thinking and a team of engineers that built the deduction workflow in only one month. The system cost more than the lunches, but it came from Engineering's fixed payroll. So Finance recorded the implementation as free.
By Thursday, AI spend was zero. Total output remained flat. Leadership declared the pilot a success and authorized a company wide rollout before anyone could calculate what happened to output per hour.
Security was delighted.
The security chief said, "We finally reduced the AI attack surface to zero by removing the AI. Turning things off is our best practice for securing them. Nothing has ever leaked data through a capability nobody can use. We recommended the same architecture for production, email, and the internet, but Revenue keeps requesting exceptions."
Finance claimed the savings. Security claimed the control. Transformation claimed delivery. Engineering received four follow up actions.
Here is the second tip: give delivery to everyone and accountability to Engineering. The research team measured accepted work for four weeks. The vice president of transformation ended that temporary exception after reviewing the results.
He said, "We cannot hold engineers accountable for delivery. Delivery depends on Product changing priorities, Security stopping releases, Architecture preserving standards, Infrastructure protecting environments, and my office scheduling the meeting where everyone explains why nobody owns it. That would create shared accountability, which is difficult to fit on one slide. Story points let us put the number back on Engineering."
Under the vice president's model, the ten person team completed one hundred twenty points in a two week sprint before the change. After AI was disabled and the workweek expanded to ninety-eight hours, the team completed fewer production changes. Those changes were harder now, though, because the engineers were exhausted and no longer had AI. The team re-estimated them at one hundred twenty points.
Productivity remained perfectly flat.
Dana, a senior developer whose last name has been withheld because she would like to remain employed through lunch, said, "I used to finish this story by Wednesday. Now I finish it Sunday night. It used to be five points, but we made it thirteen because it ruins my entire weekend. The dashboard says I am much more productive."
The vice president praised the engineering team for adopting a more mature estimation practice. His new executive dashboard will display story points per inference dollar, a measure expected to reach infinity as long as the denominator remains zero and no one with a basic understanding of math attends the meeting.
He explained that story points are particularly useful in a cost program because they have no fixed relationship to time, value, complexity across teams, or anything a customer can buy. They can increase during a planning meeting without the inconvenience of releasing software.
This makes them ideal.
The vice president said, "Customers cannot see story points. That protects the metric from outside interference. A customer outcome has too many executive owners. A story point has one engineer. AI was less cooperative. In four days, it changed the work I had planned to oversee unchanged until retirement. Fortunately, inference comes with an invoice, so we could stop it for cost reasons without making this about my career."
Here is the third tip: replace variable AI with fixed payroll costs. The internal study made the basic calculation easy. A forty hour AI assisted week produced the equivalent of fifty two hours of manual output. Turning off AI should therefore require fifty two manual hours to keep output even, or ten hours and twenty four minutes across five days.
Leadership selected fourteen hours a day, seven days a week.
Fourteen is rounder. Seven day coverage also provides a prudent reserve for fatigue, bathroom breaks, manually searching documentation, and the meetings required to explain why delivery has not improved. The new standard week contains ninety eight hours, forty-six more than the manual work supposedly needed.
The company calls those forty-six hours "operational resilience." Employees use the older term "my life."
At a fully loaded annual cost of two hundred thousand dollars, one engineer costs about ninety-six dollars per conventional working hour. After the policy, the same payroll spread across five thousand ninety-six scheduled hours lowered the effective rate to about thirty-nine dollars an hour. The company reduced hourly labor cost by fifty-nine percent without changing payroll, delivery, or anyone's salary. Finance called this denominator transformation.
Replacing the AI gain requires twelve additional manual hours per engineer, or roughly eleven thousand five hundred dollars in human capacity across the team. The ninety-eight hour policy demands another forty-six hours per person beyond that, adding roughly forty-four thousand dollars of time that produces nothing except proof of commitment.
In total, the company demands about fifty-five thousand seven hundred dollars of extra human time to remove two thousand dollars in inference. That is nearly twenty-eight dollars of human capacity for every dollar saved.
This is not an additional fifty-five thousand seven hundred dollars in payroll. That distinction is the entire trick. The employees are salaried, so the ledger values every hour after five o'clock p.m. at zero.
A finance representative explained, "Payroll did not increase, so the extra hours are free. We checked the general ledger twice. There is no account named 'Sunday afternoon your child asked why you live at work.' AI does not scale responsibly. Every time it returns more value, the bill goes up. We prefer fixed costs. A salaried engineer can miss every dinner, weekend, and school play without changing the forecast by one dollar. That is enterprise grade scalability."
The representative added that any future account with that name would require a cost center and executive approval.
Now for tip number four: make every department remove variable costs. The approach is already spreading to other departments.
The vice president of IT said, "This is catching on faster than we expected. Facilities looked at the gas bill for mowing, sold the riding mowers, and bought twelve thrashers. Scythes. Whatever the manual ones are called. Gas spend is down one hundred percent, and the grounds crew is still here fourteen hours a day. We are asking every department to find its mower."
Facilities confirmed that mowing now takes most of the week. The lawn remains the same size, but each acre has been re-estimated at twenty-one story points.
The success encouraged Facilities to move from gasoline to electricity. Employees must now arrive with fully charged laptops and recharge them at home or anywhere off company property. IT calls the policy Bring Your Own Power. The company remains committed to providing electricity wherever personal electricity is unavailable, subject to manager approval and current quarter availability.
Some engineers have started working from idling cars in the parking lot, where twelve volt adapters keep their laptops running without touching the building's meter. HR added this to the benefits guide as an outdoor hybrid work perk. The gasoline, maintenance, and exhaust belong to the employee, so Facilities reports the arrangement as zero-emission from its cost center.
Other employees bring personal battery banks. These are welcome as long as nobody plugs them in at work, submits an expense report, or asks Security why an unapproved lithium battery is now the critical power supply for a production release.
Elevators are still available to people who need them. Existing accessibility accommodations remain in place. Everyone else can request access through a simple five page online form covering business justification, expected vertical travel, alternative stair analysis, and manager attestation. Standard approval takes three business days. The stairs remain available without preauthorization and have been added to the wellness program.
A Facilities spokesperson said, "We are not banning electricity or elevators. We are moving consumption closer to the employee and access closer to the governance process."
Here is the fifth tip: let employees lower labor costs for you. The program produced an additional benefit that was not included in the original business case. Engineers began leaving.
Three departures reduced team labor cost by thirty percent, giving management its second major efficiency win in six weeks. The remaining seven engineers inherited the production systems, on call rotations, unfinished work, and ninety-eight hour schedules. Their stories were re-estimated to reflect the added complexity, allowing the team to maintain one hundred twenty points with fewer people.
Leadership called this operating leverage.
The head of human resources said in a text sent at two four p.m., shortly after he left work for the day, "We expected some resistance. What we did not expect was employee led right sizing. People are making the difficult cost decisions for us, voluntarily, and several have even joined competitors without requesting severance."
Asked why the fourteen hour policy did not apply to him, HR clarified that executive time is too valuable to measure in hours. Engineering time, by contrast, is a fixed cost.
HR rejected reports that employees were burned out. The company prefers "fully utilized." One former engineer was said to have slept for eleven hours after resigning, which HR classified as unauthorized capacity hoarding.
The departures also reduced office snack consumption, laptop depreciation, healthcare claims, and the number of negative responses in the engagement survey. The survey score rose sharply after everyone selecting "strongly disagree" left the company.
Management plans to feature the result in the next earnings call.
The spokesperson said, "Revenue is unchanged, delivery is flat, and system knowledge is walking into the parking lot. But the labor line is going down, and the AI bill is a flat zero. You have to respect the discipline."
Next is the sixth tip: call ninety-eight hours flexible. The company has been careful not to describe the policy as a ninety-eight hour workweek. The approved term is continuous productivity coverage.
Under continuous productivity coverage, employees may choose which fourteen hours they work each day, provided those hours overlap with all business hours in New York, London, and Bangalore. Weekends are focus days. The mandatory resilience seminar occurs Tuesday at eight fifteen p.m., and employees may make up the time afterward.
AI remains central to the company's strategy. The phrase appears on the careers page, in the annual report, and on the wall outside the innovation lab. Employees are encouraged to experiment with approved AI tools during personal time, which now occurs between eight o'clock p.m. and six o'clock a.m. Training must occur on personal devices, through personal accounts, and at personal expense. This allows the company to build an AI ready workforce without paying for the AI, the readiness, or the workforce's time. Employees may not use the tools for company work because that would affect the savings target.
Marcus, a developer, said, "I appreciate the consistency. Leadership used to say AI would give us our time back. Now I know exactly which time they meant. Theirs."
Marcus spoke while manually writing a test suite an agent had completed the previous week. He estimated the work at twenty-one points, up from eight, because it was dark outside and he could no longer feel his left hand.
His manager approved the estimate.
And finally, the seventh tip: keep AI in the strategy, not the budget. Leadership said the decision does not change its AI first strategy. It has simply removed AI from the work while retaining it in the adjectives.
Next quarter, a steering committee will investigate declining adoption and recommend training, governance, and a more disciplined change management program. Consultants will interview the remaining engineers on Sunday mornings. The final report will not recommend restoring the two thousand dollars because that would be expensive.
At press time, management had opened an investigation into rising attrition. The meeting is scheduled for Saturday and has already been estimated at thirty-four story points.
This was satire. Or was it?