Transcript
How Nordic leadership stops AI sabotage.
Full transcript of the podcast episode.
- 01. One in Three Are Sabotaging AI
- 02. Change Done To People, Not With Them
- 03. The Public Contract With Your People
- 04. Expansion, Not Subtraction
- 05. Scoping One Narrow Win
- 06. The Usage-Metric Trap
- 07. Champions Who Fail Out Loud
- 08. The Harsh Ground Truth Before Scaling
- 09. The Human Edge
One in Three Are Sabotaging AI
Host A
A third. Like one out of every three. It's a pretty crazy number to start with, isn't it?
Host B
It really is. That is the global statistic for how many employees admit to actively sabotaging AI at work.
Host A
Yeah. And we're not talking about people just, you know, passively ignoring a new software update.
Host B
Right. Exactly.
Host A
We aren't talking about them just grumbling about it in the office kitchen while they make their coffee. We are talking about active, deliberate resistance. Which is wild.
Host B
Totally.
Host A
And if you are a manager or a business professional listening right now, especially if you're navigating the highly collaborative Scandinavian market and you lead a team of more than 50 people, some of them are quietly on that list right now. And the really unsettling truth is you probably cannot name them.
Host B
It really forces a genuinely staggering reality check, doesn't it?
Host A
Oh, absolutely.
Host B
Because organizations spend months, sometimes years, and like millions of euros just fixating on the capabilities of the technology itself. The processing speed, all that stuff.
Host A
Exactly.
Host B
The entire focus is on model integration, cloud architecture, whatever. And in doing all that, the human variable, the actual people who have to operate these systems, gets completely pushed to the side.
Change Done To People, Not With Them
Host A
Okay. So let's unpack this. Because when I read that statistic in our sources for today's deep dive, this image immediately popped into my head.
Host B
Oh, yeah. What was it?
Host A
Well, it's like a shipping company just bought this massive state-of-the-art engine, right? They paid an absolute fortune for it.
Host B
Right.
Host A
They bolted it into the hull of their flagship, but they just cannot figure out why the ship isn't moving any faster.
Host B
And why isn't it?
Host A
Because the crew is quietly down in the hold, siphoning the fuel out of the tanks, because they are absolutely convinced that this new engine was designed to eventually replace them.
Host B
That is a perfect analogy. The tooling, it turns out, is the easy part.
Host A
Yeah.
Host B
The people are the hard part.
Host A
So our mission for this deep dive is to explore exactly why these AI rollouts fail and how leaders can actually successfully navigate this human element.
Host B
Right. And specifically using a distinctly Nordic leadership approach.
Host A
Exactly. So going back to your engine analogy, the key word there is convinced. Because resistance doesn't just, like, materialize out of thin air.
Host B
It usually isn't born out of a hatred for technology either.
Host A
No, of course not.
Host B
But it's about how the change happens. When people feel that massive structural change is being done to them rather than with them, their defense mechanisms just kick right in.
Host A
Yeah. That makes total sense. And what our source material points out is that in a Nordic or Swedish workplace, this dynamic isn't just a minor hurdle. It is absolutely fatal to the project.
Host B
Because of the culture.
Host A
Right. Exactly.
Host B
The entire Scandinavian business model runs on trust. It runs on change being negotiated.
Host A
Right.
Host B
So if a leader violates that trust by imposing a massive shift from the top down, you know, without any consultation, the rollout is effectively dead on arrival.
The Public Contract With Your People
Host A
So the author of our source lays out a very clear operational reality here. If the core problem is this deep-seated existential fear of being replaced, you basically cannot ask an employee to log into a new generative AI tool until you neutralize that fear. You have to address it head on.
Host B
Right. It requires a public contract with your people. And here we are looking at concepts like MBL, the Swedish Co-Determination in the Workplace Act, and the Samverkan or Collaboration Model.
Host A
Yeah. And for anyone listening, you know that these aren't just HR buzzwords.
Host B
No, not at all. They are mandatory cultural operating systems. They are the legal and cultural bedrock of how work actually gets done in that region. Because change only lands successfully when people feel their input shapes the outcome.
Host A
Exactly. A leader has to stand up in front of the organization and state out loud, in public, exactly what they will and will not do to their people during this transition. You have to lay out the boundaries of the sandbox before you ask anyone to play in it.
Host B
That's a great way to put it.
Host A
And, I mean, we really have to acknowledge that the employee's fear is entirely rational.
Host B
Oh, completely rational. Because if you are an employee sitting in an office in Stockholm or Oslo today, you're reading the exact same global news as everyone else.
Host A
Right. You see the constant drumbeat of American headlines about tech founders cutting thousands of jobs specifically to pivot their budgets toward AI.
Host B
Yeah. You read that and your brain immediately translates those foreign headlines into your local context. You hear the acronym AI and you instantly think, oh, more organization.
Host A
Reorganization.
Host B
Exactly. You start calculating who is going to get sidelined, who gets forced into, you know, uncomfortable retraining or whose department just evaporates entirely.
Host A
And the really tricky part for leadership in that moment is realizing that logic or broad macroeconomic data will not diffuse that emotional anxiety.
Host B
No, people don't care about the macro data when their job is on the line.
Host A
Exactly. A CEO might stand up in a town hall and say, look at the national data. There's no consistent evidence that AI is driving mass unemployment.
Host B
Or they might point to the strength of Swedish labor union protection.
Host A
Yeah.
Host B
But a macroeconomic statistic does absolutely nothing to soothe the very personal late night worry of a mid-level financial analyst.
Host A
Right. The person who thinks their specific spreadsheet forecasting job is about to be automated out of existence.
Host B
Exactly. The elephant in the room has to be addressed directly at the level of the individual's livelihood.
Host A
OK. So the obvious question here is, how does a leader actually calm those hyper-specific fears without making impossible permanent promises? Because you can't just guarantee the market will never change.
Host B
Right. You'd be lying.
Host A
But our source suggests a pretty brilliant pivot that perfectly respects the Nordic consensus culture.
Host B
Oh, the hiring freeze concept.
Host A
Yes. You commit to keeping the company lean by slowing the pace of hiring rather than cutting the people who are already in the building.
Host B
I love that. It is an absolute masterclass in change management.
Host A
It really is.
Host B
Because it's entirely honest about the need to manage future headcount and control costs, but it immediately removes the guillotine hanging over the people currently sitting in the room. The message basically becomes, you are safe. We are going to aggressively grow our capabilities and our output, but we are going to do it without radically growing our numbers.
Expansion, Not Subtraction
Host A
This is where it gets really interesting, because the text ties this directly to a philosophy from Jensen Huang.
Host B
The CEO of NVIDIA.
Host A
Yes. I mean, he is running one of the most valuable companies on the planet, heavily driving the AI revolution. And his stance is that he expects enormous productivity gains from his engineers, but he isn't letting go of any of them.
Host B
Right. He actually went on record stating that leaders who cut staff to implement AI simply lack the imagination to figure out how to use their teams in the AI age.
Host A
Which is such a powerful statement. That framing fundamentally rewires the psychology of the rollout for the entire company.
Host B
Because it shifts the focus.
Host A
Exactly. Instead of the conversation being anchored in subtraction, like what tasks are being taken away, whose budget is getting cut, it becomes a conversation about radical expansion. So the question changes to, what massive, complex projects can we finally go after now that we have this technology doing the heavy lifting?
Host B
Yes. The leader's primary job is to paint that wider vision, positioning the AI not as a replacement worker, but simply as the productivity engine underneath human ambition.
Host A
Okay. So what does this all actually mean in practice? Because I have to push back on this narrative slightly.
Host B
Okay. Okay.
Host A
If you read the financial press right now, the entire pitch for AI, the thing driving the stock market frenzy, is exactly that. Saving money on headcount.
Host B
That's true.
Host A
Wall Street seems to be rewarding companies that use AI as an excuse to trim their workforce. So how do you reconcile that?
Host B
Well, Wall Street rewards short-term margin improvement, sure. But relying on AI purely as a cost-cutting mechanism reveals a fundamental misunderstanding of how the most successful, durable companies are actually deploying the technology.
Host A
How so?
Host B
Think about the mechanics of it. If you use AI to fire 10% of your staff and your company's output remains exactly the same as it was yesterday, you haven't actually innovated your product.
Host A
Or improved your customer experience.
Host B
Right. You've just performed a financial parlor trick to shrink your cost base.
Host A
That makes a lot of sense.
Host B
The real sustainable bottom line value comes from taking your existing highly knowledgeable staff, equipping them with these models, and suddenly doing 50% more work.
Host A
Or tackling problems that were previously out of reach because they were too labor intensive.
Host B
So when a leader clearly communicates that AI is a tool for empowerment and expansion, not elimination, that dynamic shifts from resistance to curiosity.
Host A
Exactly.
Host B
But, you know, even if a leader successfully neutralizes the fear of being fired, why would the employees actually change their daily habits?
Host A
That's the million-dollar question.
Host B
Right. You still face the massive hurdle of getting them to actually use the tools.
Host A
Yeah. You cannot just blast out a company-wide email declaring, we are an AI company now, everyone start doing AI, and expect anything to happen.
Host B
Oh, absolutely not. That kind of grand corporate proclamation fails on two distinct fronts.
Host A
What's the first one?
Host B
First, the employees simply do not believe it. It sounds like every corporate jargon designed for a press release.
Host A
Very true. And the second...
Host B
Second, even if they do believe leadership is serious, an abstract statement offers absolutely no operational reality.
Host A
Right. It doesn't mean anything for their day-to-day.
Host B
Exactly. It doesn't tell an accounts payable clerk how to process an invoice faster, and it doesn't tell a software developer how to write a more secure line of code.
Host A
It means nothing for their average Tuesday.
Scoping One Narrow Win
Host A
And this is exactly where the structural realities of the Scandinavian business landscape offer a massive built-in advantage.
Host B
Yes, the environment is primed for it.
Host A
Because we are talking about regions with lean, agile organizations and incredibly flat hierarchies.
Host B
Plus, you have an unusually high level of digital maturity across the workforce.
Host A
And remarkably clean data infrastructure compared to many global markets.
Host B
Oh, for sure. And a population highly proficient in English, which is crucial since most of these LLMs are primarily trained on English data sets.
Host A
That combination is a tremendous head start. But, you know, leaders often squander that advantage.
Host B
By trying to do too much at once.
Host A
Exactly. They aim for some massive, company-wide moonshot project to prove that their AI investment was worth it.
Host B
And the source material is adamant that this is the wrong approach.
Host A
So no moonshots.
Host B
Right. You should not make your first foray into AI some unsolved, highly complex research problem that takes 18 months to deploy.
Host A
The strategy has to be remarkably narrow. You have to pick one specific, highly defined target that actually moves the bottom line.
Host B
It has to solve a real existing problem.
Host A
Whether that is driving new, measurable revenue or significantly cutting internal tooling costs.
Host B
The text highlights a couple of proven starting points, actually.
Host A
Yeah. Customer service is a big one.
Host B
Customer service is a brilliant workflow. You route the highly repetitive, routine cases to AI agents, while the complex, emotionally sensitive cases stay with human operators.
Host A
And the AI works in the background, right?
Host B
Yeah. It's working the CRM and searching policy records, surfacing relevant information onto the human operator's screen so that every conversation is faster and more accurate.
Host A
That's huge.
Host B
Another common and highly effective starting point is within engineering and development teams.
Host A
Developers usually want to play with the new toys anyway.
Host B
Exactly. They actively want to learn AI-native ways of working, like using co-pilots for boilerplate code generation or automated testing.
Host A
But regardless of whether you choose customer service, engineering, or marketing, there is a massive trap hiding in how leadership evaluates the success of that narrow project.
The Usage-Metric Trap
Host B
Oh, the usage metrics trap.
Host A
Yes. The usage metric is such a dangerous illusion.
Host B
It really is.
Host A
I like to think of it like a company buying a lavish corporate gym membership for all its employees. And leadership calls it a massive success just because the data shows hundreds of people swiping their key cards at the front door.
Host B
But are they working out?
Host A
Exactly. If you look inside, nobody's actually lifting any weights. Nobody is running on the treadmills. Nobody is getting any healthier. They are just swiping in, grabbing a free towel, and leaving.
Host B
That analogy perfectly captures the exact failure mode of early AI rollouts. Our sources point to a highly visible historical example of this from Uber.
Host A
Oh, the Uber story is fascinating.
Host B
It's wild. When Uber first began rolling out generative AI internally, leadership pushed usage heavily across their workforce. They practically mandated that employees engage with the models.
Host A
And what happened?
Host B
Well, usage metrics spiked astronomically.
Host A
Everyone was prompting the AI. Mission accomplished, right.
Host B
You'd think. But unlike traditional software, where clicking a button a thousand times costs the company essentially nothing, LLMs operate fundamentally differently.
Host A
Right. Because of the compute cost.
Host B
Exactly. Every single time an employee sends a prompt to an AI model, it requires processing power on massive, expensive GPU servers.
Host A
It costs fractions of a cent, but when thousands of employees are doing it constantly for trivial tasks...
Host B
Those fractions multiply rapidly.
Host A
Yep. Uber's compute costs went through the roof.
Host B
And the fallout from that is so damaging to leadership's credibility.
Host A
Uber had to publicly backtrack, and there was talk of having to introduce strict token budgets just to stop their own employees from using the very tool they had just been ordered to use.
Host B
It's just massive whiplash for the workforce.
Host A
Talk about severe whiplash. Use it! Use it! Innovate! Wait, stop. You're costing us too much money in server fees.
Host B
Because it instantly makes the executive team look like they have absolutely no grasp of the technology they are championing. And that kind of uncoordinated reversal erodes the very foundation of trust we established as the prerequisite for a successful rollout.
Host A
Because usage is ultimately a vanity metric.
Host B
Precisely. If a pilot project's success is defined merely by the number of logins or the volume of prompts generated, the strategy has already failed. True success must be evaluated on whether the fundamental way work is done has actually improved.
Host A
And whether the end customer actually felt a positive impact.
Host B
Did the resolution time drop? Did the code ship with fewer bugs? That's what matters.
Champions Who Fail Out Loud
Host A
But making that tangible change happen requires a specific type of leadership. Especially in a flat Swedish hierarchy, you cannot rely on a steering group of directors and VPs sitting in a boardroom discussing the theoretical benefits.
Host B
No, it will never translate to the floor. You need a passionate, dedicated champion in middle management.
Host A
Middle management is where the theoretical meets the operational. If the team-level leader, the person assigning the daily tasks, isn't genuinely excited and pushing the integration, the new workflows will never take root.
Host B
But beyond just being a champion, those leaders have to actively walk the walk.
Host A
Oh, for sure.
Host B
In a culture that rejects heavy-handed, top-down mandates, employees are watching their managers closely. They need to see their leaders visibly using the tools and, honestly, more importantly, failing with the tools.
Host A
That vulnerability is key.
Host B
A manager needs to be able to stand up in a team meeting and say, Hey, I tried to use our new AI agent to write the Q3 financial summary yesterday, and it completely hallucinated half the revenue numbers.
Host A
Right. And for anyone listening who might be unfamiliar, hallucination in this context is when the AI confidently strings together words or data points that sound entirely plausible but are actually completely fabricated by the model.
Host B
It's such a weird phenomenon.
Host A
It is.
Host B
But admitting that kind of failure normalizes the learning curve. It shifts the perception of AI from a flawless, intimidating oracle that might take their jobs to a slightly unpredictable tool that requires human oversight and patience to master.
Host A
AI transformation is deeply personal. It requires behavioral change just as much as organizational restructuring. If leadership isn't participating in the friction of that journey, the team will simply revert to their old, comfortable ways of working.
The Harsh Ground Truth Before Scaling
Host B
Which naturally leads us to the post-pilot phase. You've run the narrow test, you've measured the real impact, and now the corporate reflex kicks in.
Host A
And usually, one of two things happens here.
Host B
Right. Either a team decides the pilot was a massive, unqualified success and tries to scale it everywhere instantly.
Host A
Or they hit a technical snag and completely abandon the technology as a hyped-up fad.
Host B
Exactly. Both of those reflexes bypass the most critical step in the entire integration process.
Host A
The pause required to find the harsh ground truth.
Host B
You have to get relentlessly honest about what actually happened during the pilot. Teams that skip this step usually persist out of sheer stubbornness. You know, telling themselves, it is AI, it is the inevitable future, we just have to force the organization through the friction.
Host A
Or they use one bad rollout as definitive proof that the technology isn't ready for enterprise use.
Host B
The source outlines a really fascinating checklist for this critical pause.
Host A
Because the questions you have to ask aren't just about the software's capabilities.
Host B
No, they are about the environment it was dropped into.
Host A
Right. Was there genuine, sustained managerial commitment behind the pilot?
Host B
Or did they just buy licenses and walk away?
Host A
Were the employees actually taught how the underlying models work? Were they fundamentally fail? And what a good output looks like in their specific daily role.
Host B
And that leads to a really profound observation from the text. Technical details always, inevitably, become people details.
Host A
You're not just dropping a new piece of software into a static work environment. You're evolving a completely new capability where human employees and AI agents have to collaborate in a shared fluid workflow. That requires a total rewiring of how information moves across the business.
Host B
It does. The perfect example of this is file formatting. The source notes that if you introduce an AI agent to summarize meetings, and suddenly all your AI memory files are stored in a format like Markdown.
Host A
Oh, Markdown. Developers love it.
Host B
They do. It's a lightweight text formatting language using asterisks and hashtags. But if you drop that on non-technical staff, you create instant friction.
Host A
Oh, definitely.
Host B
If a marketing manager is used to writing in beautifully formatted Microsoft Word documents, and suddenly they are handed a raw Markdown file from the AI, they can't easily read it.
Host A
They're going to panic a bit.
Host B
They will immediately ask, wait, how do I edit this? Do I still write my weekly updates in the company wiki, or do I put them here?
Host A
You change a simple technical output, and you instantly break a human's established workflow.
Host B
It creates unnecessary paralysis, and that paralysis is exactly what you must avoid when you finally decide to scale the technology beyond the pilot group.
Host A
When you scale, you have to do it deliberately, one team at a time. And you must lead the internal narrative with the concrete customer impact and the business case you proved during the pilot.
Host B
What you absolutely must avoid is drowning your non-technical staff in discussions about back-end architecture, token limits, or context windows.
Host A
Most employees do not care how the engine works. They just want to know how to steer the ship.
Host B
But there is one area where you do have to get highly technical and explicit with the wider team.
Host A
Especially operating within the Scandinavian market.
Host B
Yes, where GDPR compliance and strict data privacy expectations are absolutely non-negotiable.
Host A
You have to plainly name your safeguards. You have to literally spell out the rules of engagement.
Host B
A leader must tell the team explicitly, like, this is exactly how we evaluate and verify the AI agent's output. This is the specific tier of customer data that the agents are explicitly blocked from touching. And crucially, this is exactly where our proprietary company data lives, and we guarantee it is not being used to train external public models.
Host A
In some fast-moving, aggressive corporate cultures, that level of granular detail might feel like bureaucratic overkill. It might feel like you are slowing the project down with red tape.
Host B
But in a Nordic context, that transparency is the absolute price of trust. If employees do not know what data is safe to put into the prompt, they will be too terrified of causing a data breach to use the tool at all.
Host A
Defining those boundaries proves to the workforce that leadership isn't just flying blind on the hype cycle.
Host B
And it reinforces that trust is the actual fuel the entire rollout runs on.
Host A
It always circles back to the people.
Host B
Always. You can have the most sophisticated models in the world, but if the human operators don't trust the system or understand their boundaries within it, the technology yields zero value.
The Human Edge
Host A
Which brings us to the ultimate existential question of this whole deep dive.
Host B
The big one.
Host A
It is the question every single employee is quietly asking themselves during these transitions.
Host B
And it's the question every leader needs to be prepared to answer clearly.
Host A
If AI is eventually going to handle all the heavy lifting, the data processing, and the routine tasks, what actually happens to us? What is left for the human worker?
Host B
It is arguably the defining question of this decade. But, you know, the conclusion drawn from the analysis in our sources is actually deeply optimistic.
Host A
Really?
Host B
Yeah. The human edge is not shrinking. It is simply shifting.
Host A
The value AI brings is incredibly real and measurable. We see it in dramatically faster code generation, rapid iteration across copywriting, and the ability to synthesize massive amounts of data in seconds.
Host B
But if an organization wants a truly professional, market-ready result that a customer will actually trust, a human is still absolutely essential. You need a human expert to strip out what we call LLM-isms.
Host A
Yes.
Host B
Those weird robotic phrasing quirks, the subtle logical leaps, or the generic hollow advice that AI tends to produce. A human must ensure the output actually aligns with the nuanced, unwritten strategy and tone of the business.
Host A
So the work isn't evaporating. It is evolving.
Host B
The work is becoming different, it is becoming harder, and in many ways, it is becoming vastly more human.
Host A
The source material provides a fascinating look at the evolution of the software engineer, actually.
Host B
Oh!
Host A
Engineers are not going to stop working just because AI can write code.
Host B
Instead, they are evolving into complex system designers.
Host A
Rather than spending their day typing out basic syntax in loops, their job will be to write incredibly robust evaluation criteria.
Host B
They will spend their time pushing the AI agents against strict security and performance standards.
Host A
Managing the architecture to ensure the code actually holds up in the real world.
Host B
Exactly. They're transitioning from typists to system architects.
Host A
And critically, this shift plays perfectly into a core strength of the Swedish and broader Nordic business culture.
Host B
Empowering small, highly autonomous teams to take real, end-to-end ownership of a product's vision.
Host A
Couldn't have said it better myself.
Host B
So if we distill all of this complexity down to a final, actionable takeaway for you, the manager or executive listening right now, it comes down to this. Protect your people first. Make that public commitment to their safety.
Host A
Start somewhere small, concrete, and real. A place where you can actually measure a positive impact on the bottom line without relying on vanity usage metrics.
Host B
And above all, be relentlessly honest and transparent about where the human edge remains essential in your organization.
Host A
Because if your people can clearly see exactly where they fit into this new picture, they will move mountains to make the transition a success.
Host B
You just cannot force a technological revolution on a frightened workforce.
Host A
You have to design the new system specifically for the humans who will be operating it.
Host B
It is a massive leadership challenge, perhaps unlike anything we've seen in our lifetimes.
Host A
Absolutely.
Host B
And it leaves me with one final provocative thought for you to mull over as you look at your own teams.
Host A
Let's hear it.
Host B
The author of our source material makes a striking historical comparison, noting that figuring out how to integrate and work alongside these cognitive models is the hardest conceptual challenge the modern corporation has faced in 500 years.
Host A
500 years.
Host B
Right now, we are relying on humans for that complex system design, that strategic evaluation, and, you know, stripping out the robotic quirks.
Host A
Right.
Host B
But if AI development continues at its current blistering pace, and these models eventually master even the complex evaluation and system design tasks we are handing to humans today.
Host A
Then what?
Host B
Exactly. What is the next uniquely human skill that you, as a leader, will need to cultivate in your workforce to stay ahead of the curve?
Host A
That is a fascinating question.
Host B
We'll leave you with that.
Host A
Thanks for joining us on this deep dive.
Listen to the full episode.
Prefer to listen? Head back to the podcast episode this transcript belongs to.