“Then?”
“Refund.”
“All?”
“Yes.”
“Even where time difference uncertain?”
“No.”
“Verify.”
“Don't perform morality.”
“Correct records.”
Leah laughed softly.
“You sound like him.”
Hattie closed her eyes.
“That's not why.”
“No.”
“I know.”
The company completed review.
Forty-one invoices had improper time.
Thirty-one were correct.
Affected clients received corrected statements and refunds with interest.
The supervisor was dismissed.
Three employees received discipline and repayment agreements.
Two had objected in messages and were not disciplined for supervisor’s instruction.
One had anonymously reported the pattern.
The company disclosed issue publicly in annual integrity report.
No dramatic press conference.
No self-congratulation.
Just:
We identified billing practices inconsistent with our standards. We corrected them and changed controls.
A journalist asked Hattie:
“Does this damage your father’s legacy?”
She looked at him.
“No.”
“Why not?”
“Because his legacy was never that people bearing his name become incapable of wrongdoing.”
“Then what is it?”
“That you build systems where wrongdoing can be found and corrected even when finding it is embarrassing.”
That quote went everywhere.
She hated that.
Hattie had inherited Roy’s dislike of becoming a sentence.
Still, it was true.
The scandal became one of the most important things that happened to Machado Technical.
Not because it destroyed them.
Because it killed the most dangerous belief inside the company:
We are the ethical ones.
Afterward training changed.
New hires were no longer told only the founder story.
They were told the billing incident too.
One instructor began orientation with:
“Every organization eventually becomes capable of the thing it was founded to oppose.”
Then paused.
“The question is whether your structure can notice.”
Good.
Not comfortable.
But good.
The Open Hangar Fund evolved too.
At first most calls involved billing pressure, documentation, and retaliation.
Then technology changed.
Aircraft maintenance became increasingly integrated with predictive diagnostics.
Machine-learning systems recommended inspections based on sensor patterns.
Software vendors marketed enormous efficiency.
Most systems were useful.
Some were excellent.
But new ambiguity appeared.
When an algorithm recommended replacing a component early, who understood why?
When it failed to recommend an inspection, who was responsible?
What if a vendor’s model was tuned to minimize false alarms for one customer and maximize replacement revenue for another?
Hattie had spent her engineering career around systems like these.
She knew the danger was not “AI bad.”
That was lazy.
The danger was authority without intelligibility.
A young technician called the fund after his company instructed mechanics:
“Do not override predictive maintenance recommendations without management approval.”
That sounded reasonable.
Except mechanics were finding recommendations inconsistent with physical inspection.
One bearing replacement recommendation contradicted direct measurements.
A technician asked to document disagreement.
Supervisor said:
“The model has more data than you do.”
Hattie heard about case.
She attended technical advisory meeting.
A software engineer defended system.
“The model incorporates millions of cycles.”
A mechanic responded:
“It cannot hear this bearing.”
The room laughed nervously.
He continued.
“I can.”
Hattie liked him immediately.