“They’re losing institutional knowledge, they’re losing skills left, right and centre.” Tech journalist Natasha Bernal join the Tech Report’s Isaac Pound to talk about the repercussion of AI layoffs and how survivors productivity falls further than AI is increasing.
ADVERTISEMENT
Layoff by AI is just a recurring problem of CEOs sucking money via stock price manipulation. The goal was never improving productivity per se but to signal "improvement" to investors or speculators for pushing up the stock price even higher, and that feedbacks to CEOs' income.
Love seeing an add for oracle AI services within 1 min of this video starting.....
This is one of the most realistic discussions I’ve seen about the hidden impact of AI-driven layoffs. A lot of conversations focus only on productivity and cost savings, but this highlights something equally important which is the loss of institutional knowledge, declining morale, burnout, uncertainty, and the long-term effects on both employees and companies themselves. The learning experience through AI4Laymans and Rohvoo makes AI concepts, workflows, practical implementation, and responsible AI adoption feel much more beginner-friendly, balanced, and easier to understand instead of just fear-driven or hype-driven discussions online.
36:52 This idea is somewhat like a policy I thought about with regard to offshoring: Companies should pay something like a tariff for 'importing' labor from cheaper markets. I'd definitely support extending that notion to AI.
I have found AI very helpful in creating more work for myself. Now I am overwhelmed with information, data, projects. It’s creating more work for the people left at the business.
I send these videos to everyone that keeps asking me about people making money in AI!
I'm 64, working continuously in software development for over 40 years, and I see a bright future for myself fixing all of the bugs created by AI.
Workers who create useful automated workflows have a lot of incentive to withhold that workflow from their employers. We will also never be able to remove human expertise from these tools because modern AI models are probabalistic, meaning that problems like hallucination are inevitable. Replacing workers is a dumb goal. Enhancing workers is possible. Lowering the stress of production is possible, but we don't even need AI for that. We just need a culture that does not demand infinite productivity growth.
Generative AI in its current form makes the most sense for small startups and owner-operators. It has the potential to create pure chaos in any organization that depends on hierarchical structure.
These companies doing this, it feels like this is going to at some point comically backfire. The technology is fundamentally flawed, and cannot think, nor process thought or reason as it's pattern matching. It's actually almost a horrible comedy because the stock market getting puffed to new highs feels like it's going to eventually comically backfire when everyone realizes this technology just isn't quite working as planned. It's the elephant in the room indeed. The more complicated the input the worse the jagged frontier of poor reasoning gets.
Same kind of productivity loss happened to the car repair industry. But through cost cutting methods on not training people. Now if you need a car repair, it takes forever. They are wrong and replace parts that didn't fix the problem.
I’m a software developer and I used AI mainly to test its capabilities and limitations, not to replace people. My conclusion is clear: AI is not a substitute for skilled technical professionals. AI should remain a heavily supervised support tool with strict limits on autonomy, not a replacement for human expertise and responsibility. Companies blindly replacing people with AI are playing with fire. If this continues, it will lead to massive economic instability, unemployment, social unrest, and eventually backlash.
2026 marks the FIRST year in the global company I'm working for (over 20K staff globally) where there are NO layoffs since before 2020. They have been pedaling "AI" HARD since 2023 and in 2025 they realized it's not working out even close to what they were expecting, so they've put on full parking brake on the layoffs now and are trying to figure it out moving forward.
shes wonderful! thanks for having her!
Natasha is an excellent resource. Informed, rational and an excellent communicator on these subjects.
I work in control systems and - beware of AI code. I know this firsthand. It generates code that compiles, yes. But it has no concept of safety. For example, it will open valves without considering pressure differential, or it will move a robot arm through a closed door. Double-check, triple-check what AI is producing.
I've used Claude for coding and it's mostly hype. If there's a problem that is obviously in its training data it can bang out the code. If it's not in the training data it's completely hopeless. Worse it's completely confident while being useless.
The problem is enhanced when decision makers aren’t only ignorant but *actively ignore* those who aren’t. I was pro-AI, but after doing some use-case tests, I realised that the product just wasn’t good for what we needed. It would save no time that wouldn’t then be lost in double-checking everything it did. The whingy, squirming pressure from my bosses to find *something* good to say about it, *some* way it could be used to save time, was intense. Very much a situation of ‘your job is to justify what we want to do, not pour cold reality onto it’. Disgusting behaviour, and very depressing for morale.
In-house self hosted open source LLMs + APIs are significantly more cost effective than corporate licensing external frontier model providers. When said companies start charging per token usage this will be the direction of travel.
A local company (In Norway) went the opposite direction and hired more people. And now they have a really popular ad campain that gets loads of people to use their app. (It's a financial advice/overwiew app), their slogan atm is "with us, you'll never talk to a chatbot. All our customer reps are human".