After the SaaSpocalypse: Software Sprawl, Fewer Managers, and Where to Steer Students

Written By J.Diamond

My honest take on where work is headed, and what to do about it

You may have heard of something called the “SaaSpocalypse.” SaaS (software as a service): think all the standard tools we use in legacy corporate companies, like Outlook, Photoshop, etc. The SaaSpocalypse was coined when, in the first week of February, nearly a trillion dollars disappeared from software stocks. Not because the software stopped working. Because investors looked at AI and asked a very simple question: if AI can build the app, then why would anyone keep paying for that app?

We are living through two changes at once, and they are moving at very different speeds. The technology is moving fast. The way we actually work, the way companies hire, organize, decide and grow people, is moving much slower.

AI can already write code, build simple tools, and handle much of the routine coordination that fills a manager's day. And it keeps getting more capable. But companies will take closer to ten years to change how they hire, organize and work, because people, budgets and habits move slower than technology. The gap between what AI can do and what companies actually do is where most of us are going to live for the next decade. So I want to walk through it in three phases. What is really happening and what it means for those of us who are not 28 and building startups.

Where we actually are right now

Start with the software. For ten years, many of us paid Adobe every month for Photoshop and Lightroom because there was no real alternative. That is quietly ending. Free tools have caught up, and AI can now help rebuild the plumbing underneath expensive software, the file formats and features that kept us locked in. Investors noticed. By spring, Adobe and Salesforce were trading at roughly half their early-2025 valuations.

Google reported this year that three-quarters of its new code is now generated by AI, then reviewed by engineers. (Pause for effect here.) The company that built much of the modern internet now has machines writing most of its software.

Then there is Microsoft, the toolkit almost every corporate office in America runs on. Word, Excel, Outlook, Teams. Microsoft now has more than 30 million paid Copilot seats. It has started charging differently, moving away from paying per person and toward paying for the work AI agents actually do. The company is preparing for a world with fewer people at desks.

And the org chart is changing too. Korn Ferry found that 41% of employees say their company has already cut layers of management. GitLab cut up to three layers of management and broke its engineering teams into 60 small, autonomous units. Block cut 40% of its workforce and published a plan to let AI coordinate the work middle managers used to do, though even that plan keeps people in charge of strategy, ethics and the hard judgment calls. For a glimpse of where this is heading, look at the new kind of companies built around AI from day one. Lovable is one of the leading examples: a vibe-coding tool where you describe the app you want and AI builds it. Elena Verna, who has worked there for more than a year, says the company has no titles beyond IC, Lead and Head, and that roles are so blurry people jump into whatever work needs doing.

Here's the distinction: A manager used to do two jobs. One was coordinating: passing information up and down, tracking who's doing what. AI is taking that job. The other was leading: making judgment calls, owning decisions, helping people grow. That job isn't going anywhere. The problem is that companies are cutting both at once.

Now the counterweight: Oxford Economics found that AI was blamed for only about 4.5% of reported job cuts. Deloitte found that only about a quarter of organizations have moved much of their AI work beyond the experiment stage. A lot of companies are cutting because of the idea of AI, not because AI is actually doing the work yet.

Both things are true. The change is real. The hype is also real. Holding both of those at the same time is the most useful skill I can offer you right now.

Phase one, 2026 to 2028: free features, expensive trust

In the next two years, almost any tool that is mostly a nice interface gets copied. Photo editors, scheduling tools, forms, simple customer databases, design templates. Features become nearly free. But trust does not.

A company can't run its client files through a free, AI-built clone of Photoshop if nobody is responsible when it corrupts something. So the big software companies survive this phase by selling what a clone can't offer: security, legal protection, and someone to call when it breaks.

Microsoft is the clearest example. Companies won't keep paying for Word because Word is special. They'll keep paying because Microsoft holds their email, their identity, and years of their documents and decisions, and keeps all of it secure. That's the new reason to stay. Not the file format. The question every company will ask is simple: is our memory safe?

Inside organizations, companies will keep cutting layers of management, and it will often be messy. The managers who remain will look after far more people than before, with less support. Many will be quietly exhausted. If that's you, you are not imagining it. Weren’t we already exhausted before AI?

Phase two, 2029 to 2032: building beats buying

This is when companies start catching up to what the technology can already do. They'll hold on to the big platform that keeps their memory safe. But on top of it, most mid-size companies will stop buying a new app for every problem and start asking AI to build what they need, when they need it. The software budget moves from dozens of subscriptions to something harder to see: AI usage, oversight, and keeping all these pieces connected.

Here is the part that worries me. We will not end up with less software. We will end up with more. This is called SPRAWL. 

Instead of two hundred subscriptions, a company may have two thousand small tools that someone built on a Tuesday afternoon and nobody maintains. No documentation. No owner. And underneath much of it, free open-source code that is itself under strain. Researchers in Europe have warned that as people build with AI instead of engaging directly with that code, the volunteers who maintain it lose the support that kept them going. And when Ed from accounting leaves the company? Who knows what agents he used? And imagine what it will be like for regular people and solopreneurs? I’m sure, like me, you already have gmail, notion, Canva, IG, Pinterest, Spotify, Youtube and on and on… now imagine adding Agents to work some of those? THAT, my friend, is more mental load than I signed up for.

Some companies are already working on this. At Lovable, every AI agent has an “agent parent”: a real person responsible for keeping it accurate, updating it, and fixing it when it starts giving bad answers. That person is usually the subject matter expert, the one who already knows the answers. And here's the part I love: when the whole company relies on that agent, the expert behind it becomes more valuable, not less.

But sprawl creates work, not just worry. Gartner predicts the average Fortune 500 company will have more than 150,000 AI agents in use by 2028, and expects a whole new category of software built to oversee and secure them. Someone has to run all of that. This next part is my own guess, not a finding. Possible job titles: an Agent Manager (the agent parent at a much larger scale), an AI Librarian (who keeps an inventory of every tool a company has built), and an AI Quality Reviewer (who checks AI-built tools before people rely on them). None of these are pure coding jobs. They're judgment jobs.

The free alternatives we are celebrating today may get shakier just as everyone comes to depend on them.

And this evolution I happen to like A LOT: the org chart changes shape in this phase, too. It stops being a map of who reports to whom and becomes a map of what people can do. Small teams form around a customer, a product, or a problem. AI handles the status updates and handoffs. Titles fade inside the company but stick around outside it, because clients, regulators and hiring managers still need to know who is accountable.

Phase three, 2033 to 2036: the bill comes due

People ALWAYS talk about the impact on jobs that AI is having. Mostly on entry-level jobs. But there is a part of this phase that I’m not seeing a lot of detail on, and it's the one I'd bet on most.

Middle management was never only about passing information up and down. It was where people learned to lead. Where you figured out how to handle a difficult employee, make a judgment call with half the facts, or tell your boss something they didn't want to hear. It was the training ground.

We are removing the rungs of the ladder to these management jobs and then wondering, later, why no one climbed up.

By the early 2030s, many companies will have experienced executives at the top and capable people doing the work, with very few in between who've learned judgment, leadership and how to sit with ambiguity. I expect that gap to become expensive. People who can lead humans and AI together, who can explain a decision, own it, and help others grow, will be rare. And rare things get valued.

For parents and college students, the research points in the same direction, with one caution. Entry-level jobs are where the pressure shows first: one study found about a 16% drop in entry-level roles for 22- to 25-year-olds in the most AI-exposed jobs since 2024. My read, not a finding: the safest bet isn’t avoiding tech or chasing it. It’s pairing AI fluency with a real field, like health, finance, law, education or operations, where judgment and context matter.

So what will companies look like by then? I think we'll see two kinds. Small teams of people who can do a bit of everything, working side by side with AI and calling in outside experts when they need them. And big companies with far fewer managers in the day-to-day work, but very clear about who answers for it when something goes wrong.

The software we use every day will start working differently, too. Today, we open an app and do the work ourselves. More and more, we'll describe what we need and AI will do it for us. Goldman Sachs expects AI agents to make up more than half of business software by 2030.

My core message

Writing this, I felt two things at once. A little dizzy at the speed of it. And strangely calm. Could adulting actually pay off?

Because when I look past the headlines, what gets cheaper is the stuff that previously mattered most: knowing the software; polishing the perfect slide deck; having the right impressive title on the right org chart. AKA all the typical corporate infrastructure.

And now what becomes more valuable is the stuff we've been building our whole careers without calling it a skill: judgment. Context. Knowing people. Spotting BS. Managing up. Emotional intelligence.

Turns out all those years of sitting through bad meetings, talking a client off a ledge, and reading the room weren't wasted. They were training. AI can hand you a hundred answers in ten seconds. Someone still has to know which one is right, and get people to actually act on it. That someone is worth more every year.

TL;DR

  • The timeline: AI can already write code, build simple tools and handle routine coordination. Companies will take closer to ten years to change how they hire, organize and work. That gap gives you time to adapt, if you use it.

  • Jobs: Entry-level and routine coordination work gets hit first. For most people, the change will be slower and more uneven than the headlines suggest.

  • Software: Tools are becoming cheap or free. Companies will pay for trust, security and their own data, and will struggle with sprawl: thousands of AI-built tools no one owns.

  • Managers: The coordinating part of the job (status updates, handoffs, reporting) is going to AI. The leading part (judgment calls, owning decisions, growing people) is becoming rarer and more valuable.

  • Where to lean in: Judgment, context, deep expertise in your field, knowing your company's data, explaining and owning decisions, and leading people and AI together.

  • For students: Learn to use AI, but don’t stop there. Pair it with deep knowledge of a field and the people in it. The first rung of the career ladder is getting harder to reach, so real-world experience early (internships, projects, part-time work) matters more than ever.

  • The bottom line: The tools are becoming a commodity. Your judgment is not.

Sources

J.Diamond
Next

Copilot Enterprise UAT