Companies are replacing you with AI. With thousands of layoffs we have seen happen recently, the narrative is that AI is replacing cybersecurity folks, engineers, coders and in certain cases even entire departments. Yes, AI is replacing a lot of what you do today.
Here’s the caveat though: Most (if not all) companies are replacing you with AI, thinking it’s a great idea. It is not. Not necessarily right now, and definitely not in the way it’s happening. What do I mean by that?
AI Layoffs and The Panic
Companies are rushing to replace engineers, coders, and cybersecurity professionals with AI. Their reasoning? AI is faster, cheaper and more efficient. But here's the truth: this approach is flawed and dangerous. Just look at what happened with Klarna. In 2022, they laid off ca. 700 employees. In 2023, they stopped all hiring as Klarna focused on building AI-powered systems and capabilities. In 2024, they announced how AI customer service agents were doing the job of 700 human employees. In 2025, they recently reversed their decision realising they made a “mistake” in prioritising cost over quality, and that they are planning a large-scale recruitment drive to bring back human workers. In 3 years, they’ve come full circle or at least are trying to.
Why is that a problem and what can you do? Let's dig in...
Welcome to The Predictability Factor by Monica Talks Cyber, a weekly deep dive and POV at the intersection of AI, Security, Privacy and Tech, written by a hacker and CISO, to help you Go From Chaos to Resilience in The World of AI.
Today’s edition of The Predictability Factor by Monica Talks Cyber, covers:
If you haven’t already, do me a favour, subscribe and help me make an even bigger impact. Let’s dig in!
The Hidden Dangers of Mass-AI Layoffs
1. Buggy Code at Scale: In May, ironically enough, the Ex-Director of AI at Microsoft got laid off. We are seeing entire dev teams getting replaced by AI. Vibe coding is cool. Until is isn’t. Vibe coding may be great, and yes it makes virtually everyone a “coder”, but 90% of that 90% code will end up being buggy, insecure and in certain cases even fatal. Most of this unsafe code will end up being used and reused without understanding any implications. Sure, engineers and coders can also produce buggy, insecure and faulty code, but reducing that threshold means two things: While on one hand, it’s amazing opportunity for accessibility and bringing the world forward. On the other, because the threshold is so low now, we need a better, faster and more efficient way to detect this buggy code, especially if it can mean insecure code becoming a part of the entire value chain of a company. AI-generated code can be riddled with bugs and security vulnerabilities.
The problem isn’t vibe coding by itself. The problem is as and when buggy, unsafe and insecure code is created, no one’s verifying it at the rapid speed at which it is being created.
Without proper oversight, vibe code will be created, used and reused at scale and speed, leading to systemic issues.
2. Overreliance on AI: Believing AI can handle all human tasks today is a myth. Human insight is irreplaceable in understanding and mitigating complex risks. We need more people in tech, cybersecurity, finance, etc. not just those that understand code, but also those that understand the psychology behind customer experience, cyberattacks, emotions that are triggered, the flawed fundamentals around “throwing money at tech” and why that doesn’t work, how to motivate stakeholders to do the right things, how to leverage AI, human skills and competencies, how to drive the business forward, how to communicate, how to build and execute a vision, how to become and influential leader, and how to help business continue to do “risky” things while bringing innovation, and managing risks, safety concerns, ethical issues and the darker side of AI. The list goes on.
3. Talent Gap Widening: Replacing skilled professionals with AI, no matter the industry, will just widens the talent gap, leaving organizations vulnerable. Just look at the example of cybersecurity. It is not just about coding, hacking or being a SOC analyst. Cybersecurity industry requires way more skills that AI can replace today, let put replace an entire department of security engineers or hackers. Sure offensive AI is definitely one of the key AI use cases that is moving faster than others. But AI outright replacing cyber folks will just increase that talent gap. Your organisation and business doesn’t become efficient by outright replacing skilled workforce with AI. Again look at the examples of Klarna, Duolingo and others.
On top of that, mass AI-layoffs will just create problems you never heard of, issues you no longer have people to fix anymore, and further problems in trying to fix AI with AI that AI already didn’t understand to begin with, which will ultimately lead to more inefficiencies.
You don't become efficient, faster and lean by doing some random and outright replacement of your entire workforce. You become efficient by utilising skilled workers for things that AI clearly cannot lead at (see point 2) and then building an augmented team of AI and people, where AI is replacing the tasks and not outright laying off departments.
ICYMI:
Predictive Cybersecurity with AI
🤯 We have been (and most of us still are) doing cybersecurity maturity all wrong. Read full story —>
How to Stay Relevant Amidst The AI Chaos
Here's a reality check. Most AI layoffs aren’t necessarily AI-related. They are “market uncertainty masked under AI”-related. Knowing what you know now, how do you stay relevant?
According to LinkedIn’s Chief Economic Opportunity Officer, by 2023, 70% of the job skills will change. One of the biggest and scarcest commodities in the near future will be real human-to-human interactions and experiences.
“Everyone in every job is gonna generally be in a new job by 2030 ’cause the skills required for your job are gonna change at a fundamental level.”
As AI enhances your career, business and life, the need and demand for real human skills like empathetic decision making, contextual problem-solving, ability to the read the room in meetings, conflict mitigation, complex problem solving, analytical thinking, ethical decision making, social influence, etc. will only increase.
Within the next 3 years, there will be so much AI, in particular AI video, people won’t know if what they see or hear is real. Which will lead to an explosion of f2f engagement, events and jobs. Those that were in the office will be in the field. Call it the Milli Vanilli effect.
The Skill Window Is Shorter Than You Think
Here is what the data shows about where the actual demand is sitting right now. More and more CISOs are going to become directly responsible for AI governance and risk management. That is not a leadership talking point. That is an open skills gap sitting inside almost every large organisation, right now, with no clear pipeline to fill it.
The reskilling noise right now is mostly surface. Learn to prompt. Use the tools. Build agents. That is the floor, and it is already crowded.
What actually protects you amidst layoffs operates at a different level.
5 key skills that will set you apart in this ever-growing era of AI, especially in cybersecurity, tech and leadership:
Analytical Thinking: Understand how AI fails, not just how it works. Most organisations have people who can deploy AI. Almost none have people who can explain why it failed, trace a hallucination back to a training artefact, identify when a model has drifted from its intended behaviour, or detect bias before it compounds through an automated decision chain. That knowledge already exists inside your current experience. You do not need to retrain from scratch. You need to reframe what you already know about systems, risk, and failure patterns, and apply it to AI.
Systems Thinking: Build your ability to connect the problem, the solution, and the architecture. The sharpest skill in the AI era is the ability to see across systems, not just within them. Architecture design is one domain where this plays out, but the real skill is systemic sight. This is rarer than any technical certification, and it is the skill the market is hunting.
Between 1985 and 1987, a race condition in the Therac-25 radiation machine's software killed patients. The engineers who built it knew exactly how it worked. What was missing was someone who could trace the software decision, to the removed hardware safety interlock, to the clinical reality of what happens when both fail simultaneously. Three teams. Three decisions. No one holding the thread between them. Today, AI systems are deployed the same way, across the same organisational silos, with the same gap in the middle. Your job is to understand how these connect to each other, find the gap between the chairs and break the silo.
Knight Capital deployed a software update that left a dormant code path active. In 45 minutes, they lost $440 million. Everyone understood their own piece of the system. No one could see the full chain. With AI integrated within enterprises and across supply chain, that reality has only gotten more complex.
In 2018, Boeing's MCAS system had engineers who understood the sensor architecture, pilots who understood flight dynamics, and regulators who understood certification. 346 people died because no one could hold the line between all three simultaneously.
That is the skill. Not depth in just one layer, but the ability to read across the multiple components, see the holistic systems picture, and see where are the cracks or where the failure lives before it completes. This skill just became even more relevant in the world of AI. According to WEF, systems thinking is projected as one of the fastest-growing skills through 2030.
Resilience Combined with Adaptability: This is the single most significant differentiator between professionals whose roles are expanding and those whose roles are being eliminated. None of these are AI-specific. All of them make you more valuable in a world run by AI systems, because AI cannot do any of them. Not holistically, not comprehensively and definitely not easily.
Negotiation: AI can analyse data and suggest optimal outcomes, but negotiation is fundamentally about reading human emotions, building trust, and finding creative win-win solutions that aren't always logical on paper. Logically, you may be doing it wrong. However, emotionally losing may give you a win-win. That’s not something AI can “reason” with. Whether across vendor relationships, third-party risk management, internal stakeholder management, crisis communication, etc. negotiation will be foundational. It is that one skill you'll need before landing your job, during interviews, in stakeholder meetings, in customer negotiations, in risk management decisions, during your board reporting, in your budgeting process, in literally your day-to-day and in everything you'll do.
Check out the full 9 Step-Framework to Negotiate Your Worth.
Human Judgement and Reasoning: This is kind of in the same category as systems thinking but a bit-more wider. AI excels at solving defined problems with clear parameters. With LLMs, they have become better at solving probabilistic problems. However, the moment the problem gets more complex, the reasoning effort declines. To add to that, complex and large-scale problem solving requires understanding interconnected systems, long-term effects and repercussions, predicting human behavior, and making decisions with incomplete information, across multiple domains. AI will get even better at a lot of this, but real human judgement and oversight will always be needed in most use cases.
Context-Based Decision Making: AI can be given business context through prompting, retrieval, or fine-tuning, but it lacks the embedded, lived understanding of organisational priorities, unwritten rules, and cultural dynamics that comes from operating within them. While we are already there to provide the model with the entire business context, it comes at a high cost, that can be financial, regulatory or other. You still wouldn’t give it everything, especially if you are in a regulated industry. While context windows of LLMs are expanding with newer models and larger parameters, they are still finite. The understanding of which "technically correct" decision might still be wrong for the business is purely a human intelligence. AI that has been trained with ethical guidelines, can only follow them within its training distribution. It lacks the situational, organisational, and relational ethical judgment that humans develop through lived experience. It doesn't understand when it is ok to bend the rules or take calculated risks, based on business context and when it may be completely illegal to do that even if technically possible.
The vicious cycle will keep spinning. It’s up to you to build your skills that help you stand out.
This newsletter is supporter by readers like you. Please share this with others and help me make an even bigger impact.
The Shocking Illusion of “The Illusion of Thinking”
This week Apple “shocked” the world. The paper revealed that LLMs don’t really “reason” or “think”. The researchers found that the 'thinking' ability of so-called 'large reasoning models' collapses when things get complicated.
But I don’t think that’s really the most shocking part. At least it shouldn’t be to. Despite that, over the last 48 hours, my social media feed’s been flooded with people calling the research “groundbreaking”.
I get it, most of us probably needed that reminder. But, here are 3 things that I see happening:
The "Shock" Fallacy Most people were shocked about an outcome that we already “knew”. That can only mean two things: Most people either didn’t really know that LLMs don’t really “think” or despite knowing that they are successfully being illusion-ed by that “illusion of thinking”.
The Complexity Collapse Even when LLMs think step-by-step, they mess up on hard puzzles. More tokens does not necessarily mean better thinking. On the other hand, LLMs also don’t know when to stop, thereby, wasting effort, compute and sometimes making things worse.
“Reasoning effort increases with problem complexity up to a point, then declines - despite having an adequate token budget.”
The Weaponisation of The Illusion Here’s where it gets even more interesting. It’s precisely this illusion of “illusion of thinking”, that is driving AI-based social engineering, incorrect decisions, fraud and other deceptive attacks, making it easier to execute and harder to detect.
The one big underlying lesson from all of this: Lack of AI literacy is real. AI literacy does not only mean what AI can do, but also what we do not want AI to do and especially what AI makes us ignore. More on that in my upcoming newsletters.
Until next time, this is Monica, signing off!
