So the temptation is to blame the economy, consumer habits, AI or Trump. They have an impact, for sure, but I think the bigger factor is interest rates. Borrowing isn't as cheap as it used to be, so big tech can't be as fun as it used be. Big Tech companies used to operate like childish nerd daycares...bright colors, toys everywhere...crazy perks, etc.
That was just the visible effects. The other effect is staff bloat and that layoffs were a bummer...so what happens to your nerds when they get old and can't keep up and decide they don't want to write code anymore? Well, they go to management. I worked at a place where you were expected to code for 3-5 years and then move to management...I had like 10 layers between me and the CEO. I only knew what about 3 of them did. The org chart was insanely complex and no one could tell me what those guys in the middle were doing. When you're competing for geniuses...if Google is laying off people regularly, they'll just go to Facebook, Netflix, or some upstart. That's been one of many reasons why top talent preferred younger companies to older ones like Microsoft.
So from 2003 to 2023, we had about 20 years where layoffs were taboo. Cruft accumulated. Companies went from nimble and innovative to bureaucratic. Take Google...they had so many awesome projects 20-15 years ago. They still have GREAT projects, which they never bother to update...for example Google Fiber or their WiFi routers...which I love, but they seem to get no love....same for their web services. I don't know for sure if that's from organizational cruft, but it would be logical to assume it is.
All those great engineers from 20 years ago who did amazing things....but now, they have kids, aren't motivated to work more than they have to and have learned the organization well enough not to get fired with cause...now they're middle managers...barely adding value, attending meetings, putting on a show, etc. This wasn't a huge issue. Interest rates were low, so you could just create a job for them and barely care if they were really providing enough value to justify their hefty salary.
But now..borrowing is expensive. Layoffs are not taboo. It's time to run your company like a real business, not a daycare for tech enthusiasts.
Let's pretend Big Tech isn't evil and isn't incompetent....I know, it's a stretch...but let's pretend Sundar Pichai hires you to run Google right...
What would you do? Organizational streamlining. You want to retain your engineers. You want to only keep managers that actually provide value. My company removed 2 intermediate managers...EXTREMELY well paid directors that only managed like 2 directors...who seemed to run everything. No one I know had any clue what those 2 did.
Now going back to AI. Having used AI since it was commercially available, it's not going to replace your engineering staff...if you know what you're doing. Yeah...my UI skills are 20 years old. I am clueless on modern frameworks. I can vibe code you a new angular app...but...how badly do you want to put something in production I don't fully understand? Are you willing to risk an outage?...a data leak? When I used AI to generate Java, a language I know well, well...it compiles about 50% of the time. Giving it the simplest problems, it solves them slightly over half the time. The solutions?...almost always dated and poorly written. Yeah, it works...but it's convoluted and basic. I am not confident it's actually saving me time. It might be. However, I spend so much time fixing their shit and confirming it actually works that I could have usually written it faster if I just did it without their help. Everyone working with LLMs knows this.
However, if your job is to summarize meetings?...to answer questions for executives?...well I can see those AI tools doing a pretty decent job. AI can be an excellent tool to act as a buffer...allowing executives to ask questions in depth before speaking to a human being.
I can see a large business like Amazon cutting several layers of white collar workers with AIs that can know all your docs and your outputs and your Jira tickets, etc....that can read repos and give you insights on progress, etc. So instead of holding 4h of meetings, you can spend 30 min asking an AI questions and condense that 4h meeting into a 1h confirm and summarize session.