The Notebook on LH401

Nora Keller had a rule for overnight flights.

Never open the laptop until the meal trays were cleared.

It was not a wellness rule. Nora had no patience for wellness rules written by people with quiet calendars.

It was a survival rule.

If she opened the laptop too early, the flight disappeared.

One email became eight.
One slide became seventeen.
One “quick review” became a private argument with a spreadsheet somewhere above Nova Scotia.

So on Lufthansa Flight LH401 from New York to Frankfurt, Nora did what she always did after takeoff.

She accepted sparkling water.
She declined champagne.
She removed her shoes without becoming the kind of passenger everyone silently judged.
She placed her phone in airplane mode.
She folded her coat carefully.
She took out the small black notebook she had carried for seven years.

The notebook was almost full.

Its corners were soft. The elastic band had lost its confidence. The pages were crowded with handwriting that began neatly on good days and collapsed into battlefield markings on bad ones.

On the first page, written years earlier in blue ink, were four words:

Promises We Made

Nora had started the notebook as a private joke.

In executive meetings, promises had a way of changing shape.

A promise made in April became a “target” in June.
A target became an “aspiration” in September.
An aspiration became “dependent on market conditions” by December.

Nora, who was then VP of Technology, had grown tired of watching commitments evaporate inside language.

So she began writing them down.

Not every promise.

Only the ones that mattered.

A customer launch date.
A platform migration.
A cost reduction claim.
A security remediation commitment.
A board-level productivity number.
A hiring freeze workaround.
An AI pilot that someone had described as “transformational” before anyone had asked who would own the workflow after the demo.

At first, the notebook was useful.

Then it became uncomfortable.

Then it became necessary.

Now, as CIO of a manufacturing software company stretched across New York, Frankfurt, Munich, Chicago, and Kraków, Nora carried it like contraband truth.

The cabin lights dimmed.

The aircraft climbed into the dark.

New York dropped away beneath them, all grid and glitter, as if the city were trying to look more certain from above than it ever felt on the ground.

Nora looked out the window until the lights disappeared.

Then she opened the notebook.

On the current page, she had written:

Frankfurt Board Review — AI Productivity

Underneath, three bullet points.

1. Board wants 18% productivity story

2. Demo works

3. Quarter does not

She stared at the third line.

Quarter does not.

It was not a sentence she could put on a slide.

It was too plain. Too dangerous. Too close to the thing everyone already knew and nobody had yet said cleanly.

The AI demo worked.

That was the problem.

If it had failed, the conversation would be easier.

They could blame the vendor.
Blame the data.
Blame integration complexity.
Blame security review.
Blame change management.
Blame timing.

Executives are very good at creating a respectful funeral for failed technology.

But the demo had worked beautifully.

In New York the previous week, the AI agent had handled a simulated customer-support workflow in six minutes. It summarized the ticket history, identified the warranty clause, drafted the customer response, suggested the next technical step, and created an internal follow-up task.

The board member from Boston had leaned forward.

The CEO had smiled.

The CFO had asked the question everyone knew was coming.

“How much productivity can we capture this year?”

That was how numbers are born.

Not from truth.

From a moment.

Someone shows a demo.
Someone powerful gets excited.
Someone finance-minded asks for impact.
Someone ambitious gives a range.
Someone cautious fails to stop them in time.

And suddenly, by the time the meeting notes are distributed, a six-minute demo has become an eighteen percent productivity assumption.

Nora had written that number in her notebook.

18%.

Then circled it so hard the pen tore the page.

Now she was flying east into tomorrow to explain why the number was both possible and misleading.

A flight attendant stopped beside her.

“Would you like dinner now, Ms. Keller?”

“Yes, thank you.”

“Still water or sparkling?”

“Sparkling.”

“Wine?”

“No, thank you.”

“Long day?”

Nora smiled.

“Long quarter.”

The flight attendant smiled back with the professional kindness of someone who had heard enough fragments of enough lives to know when not to ask more.

Dinner arrived on a tray that was more organized than Nora’s operating model.

Small salad. Warm bread. Salmon. Potatoes. A dessert she would pretend not to eat and then eat entirely while reviewing the third slide.

She ate slowly.

Across the aisle, a man in a dark sweater was already asleep, mouth slightly open, one hand still resting on a closed laptop. Nora envied the biology of strangers.

She turned the notebook page.

The next page had a hand-drawn map.

Not a technical architecture map.

A work map.

New York.
Frankfurt.
Munich.
Chicago.
Kraków.
Bangalore.

Arrows everywhere.

Customer escalations from New York to Frankfurt.
Product decisions from Munich to Kraków.
Engineering dependencies from Kraków to Chicago.
Data cleanup from Bangalore back to Frankfurt.
Security approvals from Frankfurt delaying customer work in New York.
AI pilots sponsored in New York, designed in Munich, tested in Kraków, and expected to reduce workload in support teams that had not been part of the design.

It looked less like a company and more like weather.

This was the North Atlantic operating problem nobody described on earnings calls.

The business crossed the ocean faster than the work could.

Executives flew between New York and Frankfurt in seven hours.

Decisions took seven weeks.

Sometimes longer, if legal was involved.

Nora picked up her pen.

At the top of the map she wrote:

The demo crosses the Atlantic. The workflow does not.

That sentence had been waiting for her.

She underlined it.

The aircraft leveled. The seatbelt sign turned off. Somewhere behind her, a baby cried once, reconsidered, and went quiet.

Nora opened the board deck.

The title slide looked expensive.

AI Productivity Update

Under it:

Frankfurt Board Review
Q2 Operating Session

The deck had been prepared by three teams, edited by two executives, softened by communications, and made visually attractive by someone who clearly had not attended any of the meetings where the actual problem lived.

The second slide was titled:

Progress to Date

It was not wrong.

That annoyed her.

They had made progress.

The AI support demo worked.
The warranty knowledge base had been connected.
The agent could summarize ticket history.
The agent could suggest next steps.
The agent could draft responses in English and German.
The agent could classify escalation risk.
The agent could generate follow-up tasks.

All true.

The third slide showed early productivity estimates.

Twelve to eighteen percent.

Nora stared at the number.

There are numbers that measure reality.

There are numbers that motivate reality.

And then there are numbers that threaten reality by arriving too early.

This was the third kind.

She clicked into speaker notes and typed:

The demo proves technical feasibility. It does not prove operating leverage.

Then she stopped.

That was the sentence.

Not perfect. But honest.

The AI worked inside the demo because the demo had boundaries.

The data was clean enough.
The workflow was chosen carefully.
The exception paths were limited.
The approval process was clear.
The human reviewer knew what good looked like.
The customer scenario was realistic, but not politically messy.

The quarter was different.

The quarter had old tickets.
Broken data.
Half-documented customer promises.
German warranty exceptions.
US enterprise escalations.
A Chicago account that always found a way to become urgent on Fridays.
A Frankfurt compliance reviewer who wanted explainability before scale.
A Munich product team that insisted the workflow should not create downstream technical debt.
A Kraków engineering team that had built the prototype but did not want to become permanent support for an AI process no business owner truly owned.

The demo had worked because the work had been made simple.

The company was struggling because the work was not simple.

Nora closed the laptop.

Not because she was finished.

Because she was becoming too angry to edit.

She opened the notebook instead.

There was a page from two months earlier.

New York AI Workshop

Below it:

• Customer support agent demo by May

• Warranty response time reduction

• Fewer manual escalations

• Support productivity visible by Q2

• Board narrative: AI leverage, not AI theater

She remembered the room.

A glass-walled conference room in Manhattan. Too cold. Bad coffee. Great view.

People had been excited.

Not fake excited.

Real excited.

That was what made it dangerous.

Everyone had wanted AI to help.

The support leaders wanted relief.
The product team wanted fewer repetitive questions.
The CFO wanted productivity.
The CEO wanted proof the company was not falling behind.
The board wanted a story that sounded like operating leverage.
The technology team wanted to build something useful for once, not another internal tool nobody used.

Nora had wanted it too.

She still did.

She was not anti-AI.

That accusation had become the laziest way to avoid operational honesty.

Nora believed AI would change work deeply.

But she also believed that companies were underestimating one brutal truth:

AI does not automatically fix a broken workflow.

It accelerates whatever workflow it enters.

If the workflow is clear, AI creates leverage.

If the workflow is confused, AI creates faster confusion.

If ownership is clear, AI helps.

If ownership is missing, AI produces outputs that nobody trusts enough to use.

If data is usable, AI can reason across it.

If data is political, stale, contradictory, or hidden inside someone’s head, the agent becomes a very confident intern wandering through a haunted house.

Nora smiled at that image.

A confident intern in a haunted house.

Too informal for the board deck.

Possibly perfect for the notebook.

She wrote it down.

The cabin was now fully quiet.

Dinner had ended. Trays had disappeared. The lights were lower. People had arranged themselves into the strange compromises of long-haul sleep.

Nora turned on the flight map.

The aircraft was over the North Atlantic.

Somewhere between the city that wanted the number and the city that would ask whether the number could survive reality.

New York had wanted the story.

Frankfurt would want the proof.

That was the corridor she lived in.

Not the glamorous corridor of premium lounges and airport transfers.

The real one.

The corridor between ambition and accountability.

Every North Atlantic executive knew it, though few said it aloud.

Westbound, leaders carried strategy, capital, customer urgency, investor pressure, and growth expectations.

Eastbound, they carried questions of governance, delivery, cost, compliance, and operational truth.

The Atlantic did not create the tension.

It simply gave it seven quiet hours to become impossible to ignore.

Nora opened the laptop again.

She deleted slide three.

Then she rebuilt it.

New title:

What the Demo Proved — and What It Did Not

Three columns.

Proved

AI can understand support context.
AI can draft useful responses.
AI can reduce repetitive human effort.
AI can support English and German workflows.
AI can classify escalation risk.

Did Not Prove

The workflow is redesigned.
The data is reliable at scale.
Ownership is clear across regions.
Compliance approves unsupervised use.
Product and support agree on exception handling.
The quarter will show productivity without operating change.

Required Next

Map the real support workflow.
Assign outcome ownership.
Clean critical knowledge sources.
Define human review points.
Create a governed AI-assisted delivery model.
Measure work reduction, not demo performance.

She sat back.

The slide was less exciting.

Good.

Excitement had done enough damage.

The next slide became:

Why the Quarter Does Not Yet Match the Demo

That one would make people uncomfortable.

Excellent.

She wrote:

The AI capability exists. The operating model to capture its value is incomplete.

Then:

Productivity is not created when AI generates an answer. Productivity is created when the organization trusts, routes, reviews, applies, and measures that answer inside real work.

That was the board sentence.

She could already hear the CEO asking if it could be shorter.

It could not.

Some truths should not be shortened just to fit a slide.

A man from the row behind her stepped into the aisle and stretched. He looked at her screen without meaning to.

“Board deck?” he asked quietly.

Nora turned.

“Is it that obvious?”

“Only because you look personally betrayed by PowerPoint.”

She laughed softly.

He was perhaps in his late fifties, with silver hair and the calm tiredness of someone who had spent decades pretending airports were normal places to work.

“CIO?” he asked.

Nora raised an eyebrow.

“That obvious too?”

“Former CTO,” he said. “We recognize the look.”

“What look?”

“The look of someone being asked to turn a successful pilot into a financial outcome by Thursday.”

Nora smiled despite herself.

“Tuesday, actually.”

“My condolences.”

He introduced himself as Martin Weiss. He was German, lived in Connecticut, and now advised industrial companies on technology transformation, which meant, he said, that he spent most of his life explaining to boards that software was not magic and to engineers that boards were not stupid.

Nora decided immediately that she liked him.

He nodded toward the deck.

“AI?”

“Yes.”

“Demo worked?”

“Yes.”

“Quarter didn’t?”

Nora stared at him.

“I’m starting to think there is only one flight and we’re all on it.”

Martin laughed quietly.

“Oh, there are many flights. Same conversation.”

He leaned against the side of the seat, careful not to invade the aisle too much.

“May I guess?”

“Please do.”

“The AI team built something technically impressive. The board saw leverage. Finance asked for a number. The business assumed the tool would reduce workload. But the real workflow runs across regions, systems, exceptions, approvals, and undocumented human judgment. Now everyone is surprised the quarter has not become the demo.”

Nora closed the laptop halfway.

“That was disturbingly precise.”

“Old pain travels well.”

She looked toward the window.

Outside, nothing was visible.

Just darkness.

Martin continued.

“Most companies think the gap is between old technology and new technology. It is not. The gap is between demo conditions and operating conditions.”

Nora opened the notebook and wrote:

Demo conditions vs operating conditions

Martin noticed.

“You keep a notebook?”

“Of promises.”

“That sounds dangerous.”

“It is.”

“For whom?”

“Depends who made the promise.”

He smiled.

“What promise are you carrying tonight?”

Nora looked at the page.

“Eighteen percent productivity.”

Martin winced.

“A number born too early.”

“Yes.”

“And you are deciding whether to defend it or kill it.”

“I’m deciding whether to tell the board it is possible but not yet real.”

“That is harder.”

“Why?”

“Because people prefer two simple lies. Either AI is hype, or AI is magic. You are offering the inconvenient middle: AI works, but the company must change to benefit from it.”

Nora said nothing.

That was exactly the truth.

AI works, but the company must change to benefit from it.

Boards liked the first half.

Organizations resisted the second.

Martin returned to his seat after a few minutes.

Nora reopened the deck.

She added one slide.

The North Atlantic AI Gap

Under it, a map.

New York: customer urgency, board pressure, productivity number.
Frankfurt: governance, compliance, operating discipline.
Munich: product ownership, domain decisions.
Kraków: technical implementation.
Bangalore: operational support and data cleanup.
Chicago: enterprise customer escalation.

Then:

The AI tool crosses this map instantly.
The work does not.

She stared at the sentence.

It was sharp enough.

She kept it.

The rest of the flight passed without sleep.

This no longer bothered her.

Some flights were not for sleeping.

Some were for telling yourself the truth before other people asked for a more convenient version.

At 4:36 a.m. local time, the cabin lights rose.

Breakfast appeared.

People woke into that fragile, confused state where everyone must remember who they are professionally before passport control.

Nora looked out the window.

The first line of morning was breaking over Germany.

Frankfurt waited below, gray and efficient, already preparing to receive decisions made badly elsewhere.

She sent the revised deck to herself, then to the CEO.

The reply came faster than expected.

This is more cautious than I hoped.

Nora looked at the message.

Then another arrived.

It is also probably right.

She smiled.

That was the closest thing to poetry her CEO had ever written.

At the airport, passengers moved quickly through the terminal. Some were connecting to Munich, Zurich, Vienna, Milan. Some were going home. Some were arriving for meetings that would begin too soon.

Nora walked through Frankfurt Airport carrying her notebook in one hand and her laptop bag in the other.

The city outside was still waking.

Her board session was at 9:30.

She had time for a shower, coffee, and perhaps one small emotional breakdown near Gate Z before becoming composed again.

In the taxi, she opened the notebook one more time.

On a clean page, she wrote:

What I will say

Then:

The AI demo worked.
The productivity number is not yet earned.
The gap is not technology.
The gap is workflow, ownership, governance, and measurement.
We can capture the value, but only if we redesign the work.

She read it twice.

Then she added:

The quarter will not improve because the demo was impressive.
The quarter will improve when the company changes how work moves.

That was the line.

Not the prettiest one.

The truest one.

At 9:30, Nora entered the boardroom.

There were pastries on the table. Nobody touched them. Board pastries were decorative, like optimism in annual plans.

The CEO gave her a small nod.

The CFO looked ready to interrogate oxygen.

The chair opened the session.

“Nora, we’re looking forward to the AI productivity update.”

Nora connected her laptop.

For a moment, the room showed her own desktop back to her: folders, deck, notes, one tiny mess of a life projected in high resolution.

Then the title slide appeared.

AI Productivity Update

She looked at it.

Then at the people around the table.

Then she did something she had not planned.

She closed the deck.

The CEO turned slightly.

Nora placed the black notebook on the table.

“I want to start with the promise,” she said.

The room changed.

People can feel when a scripted update has been abandoned.

They become alert, suspicious, occasionally human.

“In New York, we showed a demo that worked,” Nora said. “It was a good demo. The technology is real. The capability is real. The opportunity is real.”

The CFO’s pen hovered.

“But the quarter has not improved at the same speed as the demo. That is not because the AI failed. It is because our operating model is not yet ready to capture the value.”

She let the sentence sit.

No one interrupted.

So she continued.

“The demo crossed the Atlantic cleanly. The workflow did not.”

The chair leaned back.

That was the moment.

Nora knew the sentence had landed because nobody asked what she meant.

They knew.

Every person in the room had seen some version of it.

The tool works.
The process doesn’t.
The pilot works.
The business doesn’t change.
The slide works.
The quarter doesn’t.

Nora opened the deck again.

This time she did not sell the future.

She mapped the gap.

Data quality.
Workflow ownership.
Human review.
Compliance approval.
Support redesign.
Exception handling.
Regional accountability.
Product feedback loops.
Operational measurement.

Then she showed the North Atlantic AI Gap.

New York wanted speed.

Frankfurt wanted governance.

Munich wanted product control.

Kraków wanted technical clarity.

Bangalore wanted operational definition.

Chicago wanted customer outcomes.

The AI agent could move across all of them in seconds.

The company could not.

Not yet.

The board discussion was harder than the one she had prepared for originally.

That was usually a sign it mattered.

The CFO challenged the productivity number.

Good.

The CEO asked what had to change in ninety days.

Better.

The chair asked which workflows should be redesigned first.

Best.

For the first time, the conversation moved away from whether AI was impressive and toward where work should be rebuilt around it.

By the end of the session, the eighteen percent number was gone.

In its place was a ninety-day plan:

Select two support workflows.
Map them end to end.
Clean the knowledge sources.
Assign outcome ownership.
Define human review.
Measure actual reduction in human workload.
Use AI where it changes work, not where it decorates it.
Report productivity only when the operating model proves it.

It was less glamorous.

It was also more likely to happen.

After the meeting, the board chair stopped Nora near the door.

“You made the update harder,” she said.

Nora nodded.

“Yes.”

“Good.”

That was all she said.

It was enough.

By late afternoon, Nora was back at Frankfurt Airport.

Another flight. Another lounge. Another coffee she did not need.

She opened the notebook.

Under the Frankfurt page, she wrote:

Promise revised

Then:

AI productivity is not a technology target.
It is an operating model result.

She closed the notebook.

For the first time in two days, she felt tired in a simple way.

Not anxious.

Just tired.

There is a difference.

Outside the window, aircraft moved in slow, disciplined lines.

One plane was preparing to cross the Atlantic westbound, carrying someone else toward New York with a deck, a number, a promise, and perhaps a private fear that the company was asking technology to solve what leadership had not yet redesigned.

Nora watched it taxi away.

She hoped whoever was onboard had a notebook.

Because the North Atlantic was full of promises.

And not all of them deserved to land unchanged.