So apparently we're spending $1.5 trillion on AI this year. I mean, that's what Gartner says, and they're famous for pulling numbers out of thin air, but even if they're off by 50% it's still an insane amount of money being thrown at technology that most companies still don't understand. But here's the weird part - some of them are actually making it work, which honestly surprises the hell out of me.
The Companies That Got Their Shit Together
Here's what's different about 2025: some companies finally figured out that AI isn't magic. The ones making money did the boring stuff first - cleaned up their data, figured out governance, and actually planned their AI rollout instead of just buying whatever OpenAI was selling that week.
Meanwhile, companies that rushed to slap "AI-powered" stickers on everything are still debugging disasters. I know a guy whose company's AI chatbot started telling customers to delete their accounts when they asked about refunds - took them down for 6 hours while they figured out the training data had somehow included sarcastic Reddit comments. Turns out AI garbage-in-garbage-out applies even when you're paying $500/month for the premium tier.
"Human-Centered" AI (Translation: Including Actual Humans)
Companies are finally figuring out that AI without humans in the loop is like a car without brakes - technically impressive until it crashes into something expensive. All the buzzword frameworks basically boil down to "maybe we should have someone who understands the business involved in this AI project."
Skillsoft's pushing some "AI-native skills platform" which is consultant-speak for "training your employees to work with AI instead of just replacing them all." Shocking revelation: involving people who actually understand your business in AI projects leads to better outcomes than just hoping the algorithm figures it out.
The Boring Infrastructure Money
While everyone's obsessing over ChatGPT clones, the real money is going into the unglamorous stuff that actually makes AI work in production. Companies are dumping millions into CI/CD pipelines, monitoring tools, and integration frameworks because they learned the hard way that running AI in production is a nightmare without proper tooling.
Turns out that building reliable AI systems requires the same boring engineering fundamentals as any other software - automated testing, monitoring, deployment pipelines, and someone who can figure out why your model accuracy drops to 12% every time the data scientists update their feature pipeline without telling anyone. Who could have predicted that?
Autonomous AI Agents (Or: How to Automate Your Way to Chaos)
The latest buzzword is "agentic AI" - basically AI that can make decisions and take actions without asking permission first. It's like giving your intern superpowers and hoping they don't accidentally refund every customer complaint or order 10,000 units of something because of a typo in the training data.
Read AI has some system that watches your meetings and tells you how unproductive they are, which is honestly more useful than most AI tools. LogicMonitor's trying to automate IT monitoring and problem-solving, which sounds great until their AI decides the best way to fix a server issue is to reboot everything in production.
Why Some Companies Are Actually Making Money
The companies seeing real ROI have one thing in common: they fixed their data mess before throwing AI at it. Shocking, I know. You can't train good AI models on garbage data, and you can't deploy reliable AI systems when you have no idea where your data lives or who has access to it.
Companies that skipped the boring data governance work learned this the expensive way - through security breaches, compliance violations, and AI systems that confidently give completely wrong answers. Turns out there are no shortcuts to doing things properly.
The Spending Acceleration Reality Check
That $1.5 trillion spending number is probably real, and it's accelerating fast. AWS would probably claim something like 1.3 million Australian businesses "adopted AI" between 2024 and 2025, which either means AI adoption is exploding or their definition of "AI adoption" is very generous - like counting anyone who used a search feature as "AI-powered."
The question isn't whether companies are investing heavily in AI - they are. The question is whether they're building sustainable competitive advantages or just burning money on consultants who promise to "AI-transform" their business with expensive PowerPoint presentations.
The Software Integration Reality
Progress Software is stuffing AI into their Telerik platform, which is pretty much what every enterprise software vendor is doing now. Instead of selling AI as a separate premium product, they're just making it a standard feature because that's what customers expect.
This commoditization is good for adoption but terrible for margins. Every software vendor is scrambling to add AI features because they're terrified of being left behind, even if their AI integration is just a fancy wrapper around OpenAI's API.
Bubble or Breakthrough?
Half that $1.5 trillion is going to consultants who promise to "AI-transform" your business with expensive PowerPoint presentations. The other half is going to companies that actually know what they're doing and are seeing real results.
The split is getting more obvious: companies with their data shit together are building competitive advantages, while everyone else is burning cash on pilot projects that sound impressive in board meetings but never make it to production.
The good news is some companies learned from previous tech adoption disasters. The bad news is for every one that's building sustainable AI foundations, there are five more trying to AI-wash their quarterly earnings with buzzword soup and shiny demos.
Honestly, I'm tired of trying to predict whether this is a bubble or the real deal. Some companies are clearly making money with AI, others are burning cash on consultant PowerPoints. Maybe the real question is whether I should update my resume to include "AI experience" or if that's jumping on a bandwagon that's about to crash into a wall.