The change narrative: why we are doing this, in words a warehouse shift believes

Between 2015 and 2019, a national government unleashed this, I mean, it was an automated computer system that was specifically designed to aggressively claw back money from its own citizens. Right. And the scale of it was just, it's hard to wrap your head around.

Yeah. It chased over $1 billion from more than 400,000 people. Wow.

And the public narrative that they used to justify this whole thing was incredibly persuasive. Like, they told taxpayers, you know, welfare recipients have been overpaid, you're owed that money back, and our new computer system is going to recover it fairly and efficiently. Which, I mean, that sounds perfectly reasonable, right? Exactly.

It was this pristine, highly polished story. The only problem was that it was a structural, devastating lie, and it ultimately cost people their lives. Yeah.

It really is the ultimate cautionary tale of what happens when an organization optimizes a story purely for public buy-in, while entirely ignoring the underlying reality of the mechanism they're actually deploying. So officially, the Australian government called this the online compliance intervention. But the people who actually bore the brunt of it eventually gave it a much darker name, which was Robo-Debt.

Robo-Debt. Yeah. Right.

I remember seeing that. And the gap between that clean public narrative of, you know, fairness and the actual mathematical reality of the system is just staggering when you look at the source documentation. Oh, absolutely.

Because the system they deployed, it took a person's total annual income, as reported to the tax office, and it just sphered it out evenly across every two-week period of the year. Every fortnight. Right.

They just averaged it out. Yeah. And then the computer automatically declared a debt wherever that artificially averaged number didn't perfectly match what the person had reported receiving in welfare at that specific time.

Let's actually, let's pause and look at the mathematical mechanics of that. Because understanding the mechanism of the failure here is crucial for anyone who is deploying automated systems today. Go for it.

So imagine a citizen who works like an intense seasonal agriculture job for three months out of the year. Okay. During those three months, they earn a pretty decent income.

But for the other nine months, they're unemployed, they're earning absolutely nothing, and they're relying entirely on legitimate welfare support to survive. Right. Which is a completely standard, totally legal scenario.

Exactly. But under the RoboDebt algorithm, the computer takes that three months of intense income and it averages it across all 12 months. Oh, I see.

Yeah. So suddenly, the system looks at the nine months where the person was completely destitute and receiving welfare. It applies that mathematical average and artificially concludes, wait a minute, you were earning money during this time.

You were overpaid. Wow. So the machine literally invented a debt at a thin air based on, well, just bad, unlawful math.

That's exactly what it did. And then the system reversed the burden of proof. Right.

The individual who is often highly vulnerable would receive this threatening letter with a formidable government crust on it, essentially forcing them to somehow disprove a debt that a computer had just fabricated. That's terrifying. More than 400,000 people were forced into this nightmare of trying to prove a negative to a machine.

But here is the detail from the source material that transforms this from just a tragic bureaucratic glitch into a catastrophic governance failure. To timeline. Yes.

The department had been formally advised in November of 2014, which was well before the heavy scaled rollout of the program. They were advised that averaging income this way did not accord with social security law. They literally had the legal advice in their hands.

They knew the fundamental premise was unlawful, but they ran it anyway for years. Because they sustained that polished narrative of fairness right up until the moment it violently shattered against the reality being lived by hundreds of thousands of people. Yeah.

You know, you just cannot sustain a communication strategy that directly contradicts physical reality indefinitely. You really can't. So by November 2019, the government was finally forced to halt the averaging method and the financial fallout was massive.

Yeah, the numbers are huge. Right. In 2020, a class action lawsuit was settled for about one point two billion Australian dollars, which included seven hundred and twenty one million dollars in direct refunds to roughly three hundred and seventy three thousand people.

And the final judgment was just damning. By July 2023, a royal commission released its final report and they explicitly labeled the entire scheme, quote, crude, cruel and unlawful. Cruel and unlawful.

Yeah. And the commission formally noted the most devastating fact of all, which is that the scheme had been a contributing factor in multiple suicides. It's tragic.

People died because the department prioritized a polished, persuasive public narrative over the harsh, documented reality of their own internal data. Which brings us directly to you listening to this deep dive right now. You are a busy professional, you're an executive, maybe a manager or a leader tasked with rolling out a new technology.

Right. You might be sitting at your desk right now finalizing the deployment of an enterprise AI system that's going to fundamentally alter how your workforce operates. And look, the stakes for you might not involve national welfare law, but the mechanism of failure is absolutely identical.

The environment changes, but the failure modes of the narrative really don't. So our mission today is executive level education on exactly how you explain an AI deployment to your organization. You're going to learn how to write the one single story of change so plainly and so structurally honestly that the people in your organization who stand to lose the most can read it, understand it, and repeat it back to you without flinching.

And we really need to set the tone immediately here. We are not having a casual theoretical chat about marketing strategies or internal PR spin. Oh.

Think of this deep dive as a Harvard Business Review masterclass mixed with the stark warnings of a trusted legal mentor who has actually watched companies tear themselves apart in discovery. Yeah. We are analyzing a strict professional governance standard here.

Exactly. In this realm, every single word matters. Zero corporate filler is tolerated.

And absolutely every claim you make to your workforce must be backed by hard documented evidence that you currently hold in your files. Okay. So the core spine of everything we're analyzing today centers around one specific document.

Let's define the key term here. The change narrative. Right.

A change narrative is the single, short, load-bearing story an organization tells about why it is deploying an AI system. It explicitly details what is changing, why it is changing, exactly who it affects, and what the organization is doing about it. So it's not the cheerful subject line of an all-staff lunch email.

No. And it is certainly not a marketing slogan. It is the exact sequence of words a shift supervisor uses to explain the new reality to their team on the floor.

It's what your employee repeats to their spouse across the dinner table when they go home. Right. It's what a skeptical journalist is going to quote in an expose, and what a union or a works council will aggressively test against reality.

And understanding exactly what that document is, I think that requires a fundamental shift in how leaders think. This is the central thesis of our analysis today. The change narrative is a governance artifact.

Yes. It is absolutely not a communications product. Yeah.

Grasping the difference between those two categories is the whole game, because they are measured by completely different, totally incompatible bars for success. How so? Well, a communications product is designed to reduce friction. It's judged by whether it successfully persuades a friendly or neutral audience to buy into a vision.

If you think back to Robodebt, the government's initial story, we are efficiently recovering taxpayer money. That was a wildly successful communications product. I'm sure it pulled beautifully.

It gave ministers fantastic soundbites. But a governance artifact is judged by a ruthless, entirely different standard. It's judged by whether it is structurally true and whether it survives hostile inspection when it's heavily scrutinized by the exact people it negatively affects.

So if an executive writes a change narrative that, say, gets the whole company excited on launch day and it drastically lowers resistance in the first month, but it conveniently glosses over the fact that 30 people in logistics are going to lose their jobs in six months, that executive might feel like a communications genius, but they have failed utterly at governance. They have failed catastrophically. Because a persuasive but false story does maximum damage to an organization.

It acts as a highly volatile accelerant. What do you mean by accelerant? Well, if a lie is clumsy and obvious and poorly written, the workforce pushes back immediately, right? And the damage is somewhat contained early on. But if a narrative is highly polished, deeply persuasive and false, it postpones the moment people trust their own eyes.

Wow. Yeah. It forces the affected people to live in a reality that their own leadership is officially denying.

And that interval just multiplies the harm and the resentment right up until the inevitable collapse. I want to try an analogy here to really test the mechanics of this, because the psychology of why leaders do this is fascinating to me. Let's say I have a smoke detector in my house.

And it keeps going off in the middle of the night because there is a smoldering electrical fire deep inside the walls. The noise is deafening. It's terrifying my family and everyone is panicking.

Now if I take a screwdriver and I open up the smoke detector and I rewire it so it stays quiet, my family instantly calms down. They go back to sleep feeling totally safe. But I haven't actually solved the fire.

I have just deliberately removed their ability to escape it. That is exactly it. By optimizing for immediate, you know, buy-in and short-term comfort, I have essentially guaranteed a long-term catastrophe.

That is precisely the mechanism at play in a corporate rollout. A changed narrative that is tuned purely for buy-in rather than for the objective truth is exactly like that silenced smoke detector. It feels vastly better for the executive team right up until the unmitigated harm arrives on the warehouse floor.

The robodebt narrative silenced the alarm for the general voting public while the fire just raged for the 400,000 citizens trapped inside the algorithm. Yes. So as a leader deploying AI, you must stop treating your change story as a marketing deliverable to be finessed.

And you have to start treating it as governance evidence that will eventually be audited. Which means we need a rigorous and flexible standard to measure it against before it ever gets published. The standard we use is a fixed four-part bar.

Every change narrative must clear all four parts simultaneously. And I want to be incredibly precise about this. This is not a scorecard where you can average out your grade.

Right. It's pass, fail on every level. Exactly.

You do not get to score an A on three of the parts, fail the fourth, and consider it a passing grade. Right. Failing any one single part of this bar fails the entire narrative.

Because each of these four parts represents a completely separate, distinct vector through which a corporate story gets a human being hurt. Okay. Let's break down the four parts then.

Part one of the bar. True. True means that every single factual claim in your narrative is perfectly accurate and strictly traceable to evidence the organization actually holds in its possession today.

So it's not just not a lie. No. True, in this governance context, does not mean technically our outside counsel says this doesn't meet the threshold for outright fraud.

It means that a reasonable person, if they were handed the exact same raw data and spreadsheets that you possess, would draw the exact same picture you are painting in your narrative. So it's about the fundamental reality of the claim. If my narrative says this new AI system will free our team from mundane tasks to focus on higher value work, I can't just say that because it sounds nice.

Correct. If you make that claim, there must be real, explicitly named, financially funded higher value work that specific, identified people on your team will actually transition into, backed by your current workforce map. And if there isn't? If the real plan is to automate the mundane tasks and then lay off the workers because there is no other work for them, claiming you are freeing them up is a direct failure of the truth bar.

Yeah. That's just a lie. Robodip failed part one the moment they launched because their core claim of fairness was entirely contradicted by the legal of those they literally already held in their filing cabinets, proving the math was unlawful.

The gap between what they knew and what they said was the definition of untrue. That brings us to part two of the standard, plain. And the source material says you evaluate plainness using something called the glance test.

The glance test is a very specific operational hurdle. It means that a person working on the floor of your organization who has absolutely no financial stake in your corporate success and who only has 90 seconds of attention to spare between shifts must be able to read the narrative, understand it immediately and repeat it back to you accurately. 90 seconds.

Yes. If a sentence in your narrative requires a second clarifying sentence just to explain what the first sentence actually meant, the narrative fails the plainness bar. Wait.

Okay. Let me play devil's advocate here. If I walk onto the warehouse floor and I refuse to use standard business terminology like strategic realignment or efficiency synergies, aren't I just patronizing the workforce? Like I'm basically assuming they are incapable of understanding complex corporate realities.

Why can't I speak to them like business adults? It is an incredibly common defense from executives, but the logic is completely backwards. The glance test is actually a vastly higher bar of communication, not a lower one. And importantly, this exact same test applies equally when you are presenting to the board of directors.

Yes. Foggy, complex jargon isn't sophisticated. It is almost exclusively used as a psychological shield to conceal decisions the author does not want to take personal responsibility for.

Jargon is a hiding place. I love that phrasing. Exactly.

When you are forced to use plain language, you are forced to state what you actually mean using words the reader can independently verify. When you deploy corporate fog, you deliberately remove the reader's ability to check your claims against reality. Because they can't pin down what you're actually saying.

Right. And people on the floor are not stupid. They have highly tuned radar for corporate nonsense.

They correctly smell concealment when they read fog. They know from years of experience that strategic operating model optimization usually means someone in their department is losing their livelihood. Yeah, plainness is not about dumbing down your vocabulary.

It is a strict honesty discipline. Perfect way to say it. Moving to part three of the bar, address to the person who loses.

The narrative must speak directly and respectfully to the people who will bear the absolute highest cost of this AI deployment. Most corporate change stories fail this silently and completely. Because they're writing for the wrong audience.

Right. They are drafted to impress the board of directors or to motivate the executives who are, you know, winning the change. And then that exact same text is simply copy pasted and handed to the people whose jobs are being permanently automated away.

And the people on the floor can tell instantly that the document wasn't written for them. I mean, it's answering a question they couldn't care less about. Like, how does this algorithm improve our competitive market position and drive shareholder value? Exactly.

It completely ignores the only real visceral question they have, which is what happens to me specifically and when does it happen? Precisely. A change narrative that cannot look the losing party in the eye and answer their most pressing question is not an honest narrative. It's simply a narrative aimed elsewhere.

It treats the affected workers as an afterthought in their own professional demise. Finally, we have part four, survivable under challenge. A governance artifact must be survivable.

It means the story holds up its structural integrity when a hostile but fundamentally fair reader pushes aggressively on it. Imagine an investigative journalist, a highly organized union representative or a works council member interrogating your claims line by line. And survivable doesn't mean no one gets mad, right? No.

Survivability does not mean the narrative is magically immune to criticism or anger. It means that when the narrative is attacked, the story concedes what is demonstrably true and stands firmly on what is defensible. It does not suddenly crumble into dust because a load-bearing claim turned out to be a total fabrication.

The source material actually provides the ultimate operational test for this four-part standard. And before you finalize your narrative, it says to try this. Imagine reading your change narrative slowly, out loud, directly to the single person in your organization who loses the most from this AI rollout.

And imagine doing this while a skeptical employment lawyer and a Pulitzer-winning investigative journalist are sitting quietly in the back of the room watching you. That is the test. If you flinch reading it, or if the words sound hollow and absurd in that high-stakes context, it is not a governance artifact.

It is just a flimsy communications product wearing a governance costume. And it will be undressed at the worst possible moment, usually during a crisis. Okay, so we have the four-part standard.

True, plain, addressed to the person who loses, and survivable. Now we need to bridge from the theory into the actual mechanics of how these narratives break down in the real world. Because failure modes have specific, identifiable names.

As a leader, you need to know what they look like on the page so you can hunt them down and eliminate them before they ever reach a human being on your team. And what is crucial to understand about these failure modes is the psychology behind them. These aren't random typographical errors or innocent misunderstandings.

They all stem from a single, predictable root cause, which is that the author of the narrative is flinching away from discomfort. The executive flinch. Exactly.

Executives are human. When faced with delivering genuinely painful news like job losses or increased surveillance, they instinctively try to protect themselves and the organization from the unpleasantness of that reality. Right, it's human nature to want to soften it.

Yes. But instead of neutrally informing the person who actually has to bear the cost of that reality, they soften the blow for themselves. Once you recognize that executive flinch, the specific errors in the text become totally predictable.

Let's expose the specific ways that flinch manifests. The first one is the euphemism. The euphemism is a word deliberately chosen by the author to soften a harsh reality until the reality itself effectively disappears from the sentence.

We see this constantly in enterprise tech rollouts. Words like right-sizing, optimizing our footprint, transitioning colleagues, or streamlining workflows. Yeah, you see those everywhere.

Right. Each of these words is a deliberate choice to replace a plain, accurate, word-like layoffs or job cuts with a fuzzy concept that hopes the reader just won't notice the impending pain. The test to catch a euphemism is actually incredibly simple.

You take the sentence and you replace the soft corporate word with the plainest, truest word available. If the sentence no longer works, or if you feel deeply, physically uncomfortable sending it, then your euphemism was hiding a decision you didn't want to own. Exactly.

If you write, we are optimizing our operating model to capture efficiency synergies, and you swap it out for the truth. We are ending 30 rolls because the new AI system does part of the current work. The plain version forces you to commit to reality.

The soft version commits to absolutely nothing and reads to the workforce as pure, cowardly concealment. And once an executive gets comfortable softening the words, the next psychological step is simply hiding the data completely. This leads directly to the second failure mode, the buried number.

Right. This is when the single figure that actually matters to the workforce is shoved down into the footnotes. It's the headcount of who is leaving, or the exact timeline of the rollout, placed deep in paragraph 9, where the reader's eye naturally slides right past it.

Or worse, the number is omitted entirely from the text, replaced by vague qualifiers like a small number of rolls. Burying a number is a direct failure of the plainness bar, but it almost always fails the truth bar as well, right? Yes, because deliberately burying a material fact is a recognized form of misleading an audience. I imagine there's a variation of this that doesn't just bury numbers, but buries entire groups of people.

Yes, the buried stakeholder. This happens constantly in AI efficiency rollouts. It's when a whole category of affected people is simply left out of the narrative.

For instance, the document might address only the permanent, full-time headcount, completely ignoring the temporary workers or agency contractors whose contracts are being terminated immediately by the new software. Oh wow. Yeah.

If your narrative proudly states, only 10 rolls will be impacted, but is completely silent on the 50 agency pickers who will also be let go on Friday, you haven't just buried a number, you have buried an entire group of human beings because they weren't convenient to your story. And the workers on the floor don't differentiate between badge colors when their friends are being escorted out of the building. Exactly.

They see the reality. But even worse than hiding the numbers is hiding the person who actually made the call. The third failure mode is the passive voice that hides the decider.

Sentences like, roles will be affected, efficiencies will be realized, difficult decisions have been made. The passive voice is a grammatical trick used to launder accountability completely out of the narrative. Roles will be affected by whom? Exactly.

Who made the difficult decisions? A person who is losing their livelihood to automation is owed the basic respect of knowing the name of the human being or the specific leadership body that made the choice. You cannot use grammar to make an organizational change sound like a random weather event that just rolled in off the coast. This connects so directly to robodebt.

The government communications specifically deliberately made the computer the actor in their narrative. They use phrases like, the system identifies discrepancies. They did that to avoid having a human bureaucrat actually accuse a citizen of fraud.

It outsourced the cruelty to an algorithm. Exactly. It made the accusation feel vast and unanswerable because how do you argue with this system? The fix for this failure mode is mandatory in governance writing.

You must name the decider in the active voice. Instead of, roles will be optimized, you write, the executive leadership team decided to end 30 roles after reviewing the Q3 productivity map. It grounds the decision in reality and it names a human body that can be asked why.

But I think the most offensive failure mode is when an executive actually tries to claim the computer is doing the victim a favor. Mode 4 is the stolen benefit. This one is particularly insidious and it destroys trust instantly.

This is when an organization claims a benefit on behalf of the exact people who only bear the cost of the change. Imagine looking an employee in the eye, an employee whose entire job consists of manual data entry that is being fully automated and saying, this new AI system empowers you to focus on higher value strategic work right before you hand them a layoff notice. It's literally corporate gaslighting.

It absolutely is. Now, the benefit of the AI system might be very real. It might be fantastic for the organization's profit margins.

It might speed up delivery for the customers and it might even be great for the surviving staff who get to use the new tools. But attributing that specific benefit to the people who are actively losing their jobs is an acute, measurable dishonesty. The workforce detects it instantly.

And they never, ever forgive it. You must attribute the benefit only to the specific stakeholder group that actually receives it. Then we have the failure mode driven by panic and a desperate need to buy peace.

Mode 5. The uncheckable promise. This is when leaders offer sweeping commitments that the organization cannot actually keep just to ensure a quiet launch day. No jobs will be lost as a result of this AI deployment.

Nothing will change for your day-to-day workflow. The AI will only ever be used as an assistive co-pilot. Let's use an analogy here.

Imagine a bridge engineer. They know a bridge is structurally weak and can't handle heavy loads. But they are getting immense pressure from the mayor to avoid traffic jams and keep the commuters happy.

So the engineer puts up a fake, official-looking sign saying the weight limit is a massive 40 tons. It works perfectly. Traffic flows.

Everyone is calm. They get total buy-in from the commuters. Right up until a 30-ton logistics truck crosses it and the entire bridge collapses into the river.

And the critical consequence isn't just the collapse. No. It's that after the collapse, no one in that town will ever trust a road sign from that engineer ever again.

An uncheckable promise destroys the credibility of every single narrative you will write for the rest of your tenure. That is a perfect encapsulation of the risk. If your internal workforce map and your budget don't rigidly guarantee that promise today, it is a massive liability.

It's a blank check that will bounce the moment reality shifts. You can only promise what the evidence rigidly supports. And finally, the sixth failure mode.

The missing what we owe you. This is a narrative that clears several bars. It might actually be honest about the change.

It names the loss plainly. But then it falls completely deafeningly silent on how the organization is going to respond. So they just drop the bomb and walk away? Yes.

If you tell someone their job is ending due to automation, but you completely fail to state the retraining budget, the redeployment options, the exact notice period, or the psychological support mechanisms, the story effectively reads as, We have decided you are obsolete and you are entirely on your own. Even a brutally hard narrative of real job loss survives scrutiny far better when it plainly, immediately states what the organization owes the affected person. To really cement how these failure modes interact, let's walk through an immersive, detailed scenario.

Let's look at a fictional company, Northgate Fulfillment. It's a massive regional warehousing operation. We have Denise, who is the senior operations manager.

Leadership has just signed a contract for a highly advanced AI shift scheduling and productivity tracking system. It is going to completely alter how the picking and packing floor is staffed, monitored, and evaluated. And Denise, to her credit, has done the hard operational homework.

She isn't operating on vibes. She has a brutally honest workforce map. She knows the exact numbers.

Right. Out of 240 floor roles currently operating, exactly 30 will be eliminated over the next 18 months because the AI optimizes the routing so well. Another 40 roles will change substantially.

Eventually, those workers will be moving off the floor and into back office oversight and exception handling, fixing the specific cases the AI gets wrong. The remaining 170 roles stay mostly the same, but they will be tracked differently. So her job today is to evaluate the draft change narrative that the corporate PR and communications team has written for the big all-hands launch announcement next Tuesday.

She sits down at her desk with the draft. Now, Denise is a smart operator. She knows she can evaluate this by reading it as a manager.

She pictures a specific employee, Ray. Ray is a 19-year veteran packer on the floor. He has survived a dozen corporate transformations, and he doesn't trust a single word that comes out of the executive suite.

Denise reads the PR draft strictly through Ray's eyes. Let's look at exactly what the PR team drafted. The text says, Northgate is excited to announce an innovative workforce optimization initiative that will empower our associates and unlock new efficiencies across our fulfillment operations.

As part of this transition, some roles will naturally be affected, and we anticipate opportunities for growth and reskilling. It sounds so familiar. It sounds like every corporate email ever written.

But when you run it through the four-part standard, it triggers alarms in every single sentence. Denise flags it instantly. Let's start with the first word, excited.

Ray is not excited that 30 of his friends are losing their livelihoods. That word was written to reflect the mood of the board of directors who are saving money. It completely fails the addressee gate.

Next, workforce optimization initiative. That is mode one, the euphemism. Ray's brain translates optimization to job cuts in half a second.

It's an insult to his intelligence. Then we hit empower our associates. That is mode four, the stolen benefit.

The new AI system strictly measures Ray's packing speed and schedules his bathroom breaks. It absolutely does not empower him, and it certainly doesn't empower the 30 people getting laid off. It fails the truth bar and the addressee bar simultaneously.

Then we get to the second sentence. Some roles will naturally be affected. This is a masterclass in failure.

It commits mode three, the passive voice hiding the decider. Naturally affected. Gravity didn't do this.

Weather didn't do this. The Northgate executive leadership team signed a vendor contract that did this. And they bury the number two.

Yes, mode two, the buried number. It vaguely says some roles instead of stating the actual known mathematical reality. 30 out of 240.

And finally, it ends with we anticipate opportunities for growth. Mode five, the uncheckable promise. Anticipate is a weasel word.

It's vague spin rather than pointing to the actual funded plan for the 40 changing roles. So Denise does the only responsible thing a leader can do. She completely scraps the PR document, opens a blank page and rewrites it in front of the communications lead.

Her rewrite says next month we're deploying a new AI system that schedule shifts and closely tracks individual packing productivity. It goes live in Q3 because the system changes our routing over the next 18 months. About 30 of our 240 floor roles will end entirely.

About 40 different roles will change to handling the specific cases the AI gets wrong. That is so much better. And she goes on.

If your role is one of the 30 ending, you will not find out in an email. You will hear it from your supervisor face to face with at least 90 days of working notice, active redeployment help within the wider Northgate network, and a specific $5,000 retraining budget we have already set aside. This automation is a real loss for our team, not a gift.

Here is exactly what we will do about it. I can practically see the PR lead wincing, saying, Denise, you can't send that. That is vastly more painful than our draft.

But Denise's response captures the entire philosophy of governance communication. It is exactly as painful as the truth. Your draft wasn't actually less painful, it was just less honest.

And Ray would have known you were lying within the first sentence. Let's look at the difference in survivability there. When Ray inevitably raises his hand at the all-hands meeting and aggressively asks, what exactly do you mean by optimization? Are you cutting jobs? Yes or no? The PR draft has absolutely no answer that doesn't look like a cowardly, embarrassing retreat.

But Denise's rewrite? She has already said 30 roles. She has already decaled exactly what is owed to those 30 people and committed to exactly how they will hear the news. Because she isn't hiding a single thing, she cannot be caught or exposed.

This brings up a vital ethical distinction that leaders often struggle with. If rigorous, uncompromising honesty is the ultimate bar here, and Denise's blunt rewrite is the gold standard, a cynical executive might ask, isn't the logical conclusion just to be as maximally blunt and brutal as possible? Why not just write, 30 of you are fired because the AI is cheaper, clean out your lockers? I mean, that statement is technically true, plain, and addressed to the loser. Exactly.

But it fails a much deeper ethical test. This introduces another core spying principle we need to establish. Honesty is absolutely necessary, but it is not sufficient on its own.

The line between candor and cruelty is care. Let's define the exact boundary of that line. Cruelty is naming the loss plainly and stopping right there.

It treats the statement, I told you the truth, as the absolute end of the organization's moral and professional obligation to the human being. It views honesty as an excuse for abandonment. Candor, on the other hand, is naming the loss plainly A and D, pairing it tightly with what the organization owes that person in the aftermath.

The retraining budgets, the extended notice periods, the active redeployment efforts. Both are truthful, but... Both approaches, cruelty and candor, are technically truthful. But only candor is defensible in a modern organization.

A story that is true but abandons the person is just a different kind of harm. And as a leader, you are legally and ethically responsible for it. I can hear an executive pushing back on this, though.

They might say, wait, if I have to include all these support details and severance packages in the narrative, are I just trying to soften the blow again? And didn't we just spend the last 20 minutes establishing that softening reality with euphemisms is a failure of governance? That is a critical distinction that must be understood. Euphemisms soften the truth. They obscure what is actually happening in physical reality, which makes the narrative structurally dishonest.

Adding your support commitments softens the impact of the truth, which is the very definition of ethical care. You never, ever soften the truth. The truth stays sharp.

The truth of the job loss remains as sharp as a scalpel. But you add the obligation to demonstrate that you are standing with the workforce in the aftermath of that sharp truth. So we have mapped out the specific failure modes and we understand that we need rigorous candor, not cruelty.

But how do we physically guarantee our narrative is actually grounded in reality before we publish it? You mentioned earlier that the strongest defense against all these bugs is that the narrative must be a strict reading of evidence, not an invention. This is where governance becomes a highly operational task. Every single claim, number, and timeline in your narrative must point directly to a specific row or a specific figure on your organization's internal finalized workforce map.

When your narrative says, this affects 40 roles in claims processing, you must be able to put your finger on the exact 40 rows on your spreadsheet. It has to be that concrete. Yes.

When you say the transition begins in Q3, it must trace directly to the official vendor deployment timeline signed by procurement. If you cannot trace a claim to hard, verifiable evidence you already hold, you cannot make the claim. Let's examine a massive real-world success story that executed this perfectly under immense pressure.

Look at BT Group in the UK in 2023. They publicly announced they would cut up to 55,000 roles by the year 2030. And they specifically plainly attributed about 10,000 of those cuts directly to the implementation of AI and automation in customer service and network management.

Now, analyze why that specific narrative survived contact with reality. Was it a comfortable, happy message? Absolutely not. It was a massive seismic shock to the workforce and the market, but it cleanly cleared the four-part bar.

It was demonstrably true. It was incredibly plain. It named the specific operational functions affected.

It provided the exact numbers, 55,000 total, 10,000 to AI. It gave the exact timeline of 2030. And crucially, it completely avoided the stolen benefit failure mode.

They did not attempt to claim this massive loss of jobs was somehow a wonderful, empowering gift to the workers. Right. They didn't issue a press release saying, we are optimizing our global footprint to empower our network associates for the future of work.

Exactly. And because of that brutal clarity, it survived intense scrutiny from powerful labor unions, from the financial press, and from the global market. It survived precisely because it was a pure, unvarnished reading of the evidence they held internally.

They told the truth up front, voluntarily, rather than having it dragged out of them under intense pressure, piece by piece, over the next five years. The unions could negotiate based on facts rather than fighting through a fog of PR spin. So if I'm an executive sitting down with a messy draft right now, how do I actually build and check a narrative to ensure it matches the BT group standard? What is the systematic process? Experts use a very specific mental model to evaluate these drafts.

They don't just read the document once and say, it looks good to me. They evaluate the draft in four entirely separate, distinct editorial passes. One dedicated pass for each part of the bar.

You must not mix them up because your brain will inevitably miss things. Let's lay out the lifecycle of that editorial process. The first is the truth pass.

In the truth pass, you read the document entirely blind to flow or tone. You are only looking for stripped, factual accuracy. For every single claim, you ask, is this structurally true and can I point to the spreadsheet right now to prove it? This pass is designed to ruthlessly catch the stolen benefits and the uncheckable promises.

Okay, and next? The next is the plainness pass. Now you change your mindset completely. You are reading as that exhausted person on the floor with 90 seconds of attention.

You hunt down and flag every euphemism, every piece of complex jargon, and every buried number. Then you move to the addressee pass. Here, you read the document strictly from the perspective of the single person who loses the most.

You ask, was this written for me? Does it directly answer what happens to me? This is where you catch the passive voice laundering and the glaring omission of what we owe you. And the final pass. Finally, the survival pass.

You read it as a hostile but highly competent challenger, a Pulitzer-winning journalist, or a class action lawyer. You push aggressively on every claim. Which sentences hold their structural integrity and which collapse under pressure? And to really operationalize that plainness pass, the source material advocates for a highly effective technique called the repeat back test.

This is such a practical, ruthless tool for leaders. It is the ultimate reality check. You take your draft and you imagine a specific, highly skeptical, affected person, not a generic, hypothetical employee persona invented by HR, but a real human being you know.

Let's use Ray, the veteran packer from our Northgate example. You read the draft pretending to be Ray with his 90 seconds of attention and his deep institutional cynicism. Then you force yourself to write down in one or two blunt sentences what Ray would say the narrative actually means using his own blunt, unvarnished vocabulary.

So if my highly polished corporate tech says, we are undergoing a strategic realignment to capture efficiency synergies and optimize our delivery vectors, Ray's repeat back is going to be something like, they're cutting jobs to save money and they're too cowardly to just admit it. Exactly. And the gap between Ray's honest, blunt repeat back and your polished corporate text is exactly where your dishonesty is made visible.

It is measurable. If Ray's summary is more honest or more accurate to the reality of the situation than your draft, your draft is a failure. You rewrite the text until the unvarnished truth and your corporate draft are the exact same thing.

Let's talk about the actual lifecycle of this document because a very common pushback from executives is that this is all just academic overengineering. They say, look, we don't need a formal, heavily vetted, change narrative governance document. We're moving fast.

We're just going to send out a quick launch email from the VP next Tuesday and call an all-hands meeting to hype everyone up. That is what we call the accidental narrative and it introduces massive legal and operational risk. Here is the undeniable reality of corporate communication.

That quick launch email becomes your de facto change narrative. If you do not sit down to write one deliberately, intentionally evaluating it rigorously against the four-part bar, you get a narrative by default. So you don't even realize you're writing one.

Right. And because that default email is almost always written by someone managing the project rollout whose primary goal is to build excitement and momentum, you get the cheerful, upbeat euphemism by default. You accidentally commit the stolen benefit.

You didn't plan to lie or deceive anyone. You just naturally reached for standard project management cheerfulness and accidentally handed it to people for whom the project is an absolute disaster. Precisely.

The medium becomes the message and the message is disastrous. So once you have deliberately written it, evaluated it through the four passes, and successfully passed the repeat back test, what actually happens to this one-page narrative? You do not throw it away after the launch meeting. It becomes a critical piece of the permanent audit trail.

You file it securely in the official project dossier right alongside the workforce map and the vendor contracts it was based on. Because the governance lifecycle of an AI rollout does not end on launch day. No, it often begins there.

Months later, or perhaps years later if things go fundamentally wrong, as we saw with the robodebt disaster, an internal auditor, a labor regulator, or a royal commission is going to open that dossier. They're going to extract the exact story the organization told the public and the workers and they're going to lay it side-by-side with the internal workforce map and the evidence the organization actually held in its possession at the time. The gap between what you formally said and what you secretly knew is exactly what brings massive companies down in discovery.

And to prevent the narrative from quietly becoming a lie over time, because AI projects are notoriously unstable and always change scope, the document actually has to state its own expiration date, right? It absolutely must. It needs a fixed date or a highly specific operational trigger for review. A sentence like, this narrative applies through the end of Q3, or this narrative is to be reviewed when the rollout reaches the logistics team in Europe.

If a narrative is silent on its own shelf life, it automatically reverts to being an uncheckable promise the moment operational reality inevitably shifts. Let me throw one last incredibly common executive dilemma at you regarding this audit trail and the demand for strict truth. What if the future is genuinely fundamentally uncertain? Okay.

Say we're deploying the AI system, but the new productivity thresholds that it will measure haven't been finalized yet. The AI vendor is still tuning the model on our data. Do I just fake confidence in the narrative to keep morale up and avoid widespread panic? Do I just say the thresholds will be totally fair and easily achievable just to get through the week? The governance rule here is a concept called epistemic honesty.

You must mark uncertain futures explicitly and clearly as uncertain. You say plainly, the productivity thresholds are still being worked out. You admit what you do not know.

And then crucially, you commit to exactly how and when they will be figured out. We will finalize the thresholds in Q4, and we will do it with active worker input before anyone's performance review depends on them. So you don't try to paper over the anxiety with false confidence? Never.

Asserting an unproven, highly uncertain future as a subtle fact just to buy calm today is a profound accuracy failure. And when that false certainty is fed to people whose very livelihoods depend on it, it is a severe ethical breach. We've covered a tremendous amount of ground in this deep dive.

From the devastating human and legal tragedy of Robodet to the strict, inflexible, four-part standard of true, plain, addressed to the loser, and survivable. We have mapped the failure modes so you can hunt them down in your own drafts. The euphemisms, the buried numbers, the passive voice, the stolen benefits, the uncheckable promises, and the missing obligations.

We've explored the crucial, necessary ethical boundary between candor and cruelty, and how every single claim you make must trace directly back to hard evidence on a finalized workforce map. And we have firmly established that this narrative is a high-stakes governance artifact that lives permanently in an audit trail, not a PR speech that vanishes into the ether once the meeting ends. Before we give you your Monday morning action item, I want to introduce a final, deeply provocative thought that connects back to this entire idea of evidence and disclosure.

Something for you to mull over as you look at your own enterprise AI deployment plans. We have spent this entire time talking deeply about how to meticulously write the narrative. But what if the leadership team chooses not to say anything at all? What if they just deploy the tool? Silence is itself a changed narrative.

And it is almost always the loudest, most destructive narrative an organization can choose. When an organization quietly deploys an AI system in the background, maybe it's analyzing performance metrics, screening employee emails, or quietly rescheduling shifts, and they let the employees discover it organically on their own, the de facto story immediately becomes one of intentional, malicious deception. The workers do the math instantly.

They think, if this tool was genuinely good for us or even just neutral, leadership would have proudly told us about it. The fact that they hit it means it is a direct threat. Silence is the absolute worst change narrative you can deploy.

It starts you from a massive deficit of trust that your organization may never fully recover from. Which brings us to the Monday morning move. The concrete, immediate action you can take this coming week to protect your organization.

This Monday morning, I want you to take the draft launch email, the slide deck, or the internal memo for your next tech or AI rollout. Take that document and run it through the repeat back test. Read it strictly through the eyes of the single employee on your floor who stands to lose the absolute most from this change.

Put yourself entirely in their chair. Give yourself 90 seconds of attention and force yourself to write down what they will actually hear in one blunt sentence using their unvarnished vocabulary, not your polished corporate terms. If their blunt sentence is more honest, more accurate, or more alarmed than your polished corporate draft, do not hit send.

Rewrite it immediately. Strip out every euphemism. Put the real hard numbers at the very top.

Name the specific human beings who made the decision. State clearly what you owe the people who lose. Keep rewriting it until the unvarnished truth and your draft are the exact same thing.

Because if you just rewire the smoke detector to keep everyone quiet, the fire doesn't go out. Eventually, the reality of the system you built catches up with the story you told. And as we saw with the tragedy of Robodead, when a polished story built on a false core finally collapses, it takes the entire organization and the people trapped inside it down with it.

Ensure your story survives the inspection.