How to Tell If a Company Is Hollowing Out

Many companies aren't unable to use AI. They're afraid to really use it.

Because real AI transformation isn't about writing emails faster or making prettier slides. It forces uncomfortable questions:

• Why does this process exist?
• Why does this meeting exist?
• Why does this approval step exist?
• Does this role create value — or just maintain a process?
• Is this middle manager making judgments — or just passing messages?
• Does this leader actually understand the work — or just occupy a node of power?

Once those questions get asked, AI stops being just an efficiency tool. It becomes a mirror — showing who in an organization truly creates value, and who is only a process node.

So the natural follow-up: What happens to companies that are afraid to really use AI? Will they disappear overnight?

My view: Most won't vanish right away. They'll enter a more subtle state first. I call it a Hollow Company.

A hollow company isn't one without revenue. It still has employees, offices, and customers. It isn't collapsing tomorrow. On the surface, many hollow companies still look fine. They're still hiring, holding meetings, doing strategy, publishing quarterly reports, running AI training, talking about transformation, still telling employees:

"We're embracing the future."

But look closer and something else is happening inside: Capable people stop believing. High-agency people go quiet. Strong performers start preparing to leave. Processes multiply. Decisions shrink. Meetings get denser. Action gets slower. AI is used as decoration — not to rebuild. The company still runs. But its ability to create the future is draining away. That's hollowing out.

The scariest thing about a hollow company isn't that it isn't profitable today. It's that it still runs — but has already lost the ability to build the future.

This isn't abstract. The software industry is already feeling the pressure. In recent years, a new word has even appeared: SaaSpocalypse. Investors worry that traditional SaaS companies will be hit hard by AI.

The term is exaggerated, of course. SaaS won't disappear overnight. Many enterprise software companies still have customers, revenue, data, compliance requirements, legacy systems, integration ecosystems, and high switching costs. They won't all collapse because one AI tool shows up. But the word itself signals something: The market has started re-pricing legacy software companies.

A 2025 McKinsey article noted that AI is pushing software business models toward generative applications, consumption-based pricing, and other new patterns — while global enterprise spending on AI applications, though growing fast, still makes up only a small slice of total software spend. In other words: the big shift hasn't fully arrived yet. But the re-pricing has begun.

Gartner predicts that by the end of 2026, 40% of enterprise applications will integrate task-specific AI agents — up from under 5% in 2025. Enterprise software is moving from "humans clicking through interfaces" toward "AI executing tasks."

For many legacy software companies, that's a serious problem. If users no longer need to click through buttons, stare at dashboards, or manually push workflows through software — but can simply ask AI to get the work done — then much of what software used to be worth must be re-proven. That doesn't mean they'll all die. It means they must answer one question: If AI can reorganize workflows, why do you still deserve to exist?

Atlassian is a concrete example. It's hardly a small company. It still has a huge customer base, mature products, and a strong enterprise foundation. But in March 2026 it announced layoffs of roughly 1,600 people — about 10% of headcount. The official line: redirect resources toward AI and enterprise sales, and make the organization faster.

That doesn't mean Atlassian is disappearing. It shows a legacy software company actively restructuring to avoid being left behind in the AI era. And it also shows: Even companies that were strong in the past must prove themselves again. Their value used to come from:

• We have tools.
• We have processes.
• We have collaboration platforms.
• We have enterprise customers.

But the AI era asks:

• Are you just a container for old workflows — or a new entry point for work?
• Do you help people manage tasks better — or can AI complete tasks directly?
• Do you only record work — or can you reorganize it?

That's the real pressure on legacy software companies.

So hollow companies aren't necessarily small or financially weak. Some are still large. Revenue still looks decent. The brand is still strong. The offices are still beautiful — but inside, hollowing may already have begun. Because the future isn't measured by how big you are today. It's measured by whether you can still create new value.

I see three broad types of hollow company.

The first is the AI-Washing Company. These companies talk a big game about AI — lots of training, policies, sharing sessions, pilots, decks, and slogans. But core processes don't change. People still meet the same way. Approvals still climb the same ladder. Nobody still makes the real call. AI is only used to write faster, summarize faster, report faster — but nobody is allowed to ask: Does this process still need to exist? These companies don't lack AI. They use AI as decoration.

The second is the Cost-Squeezing AI Company. These are more aggressive. They really do use AI — but mainly to cut costs, not to rebuild the organization: layoffs, cost pressure, higher output targets, fewer people doing more with AI. On the surface, this looks advanced, because they are using AI. But the underlying logic hasn't changed. AI just makes the old system run faster.

The problem: if a company treats AI as cheaper labor instead of rethinking work itself, it gradually turns colder. Employees aren't amplified — they're squeezed. The organization doesn't get smarter; it just gets better at extracting efficiency. Short-term profits may improve. Long-term, it loses the people who actually create.

The third is the Business-Hollowing Company. This one is more dangerous, because the problem isn't slow management — it's that the company's original business value is being redefined by AI. It may have been very successful because it turned an old workflow into software, spreadsheets into systems, communication into platforms, approvals into processes, collaboration into boards. All of that was valuable, because people needed an interface to organize work.

But the AI-era question is: If AI can understand tasks, organize information, generate plans, push workflows forward, and track status directly — do users still need all those interfaces? If a product's main value is just an interface and a workflow, AI will keep making that value cheaper.

Of course, enterprise software isn't only an interface. It also has permissions, audit trails, compliance, data migration, customer support, third-party integrations, historical data, procurement cycles, and accountability boundaries. Those things still matter. So legacy software won't vanish instantly. But if a company's core value is only "we have a system" — not "we own irreplaceable data, ecosystems, trust, and accountability boundaries" — the AI era will keep asking: What do you have that can't be regenerated?

Some companies, though, become more promising in this era. They aren't protecting old workflows. They're building infrastructure for future AI builders, solo operators, small teams, and new kinds of organizations — companies that help people build products faster, deploy systems faster, organize data faster, and connect AI capabilities faster. They may not be the biggest companies today, but they're paving the way for new productivity. That's a direction worth its own essay later.

So how does an ordinary person tell if their company is hollowing out?

I think you can watch for a few signals.

Signal one: more meetings, fewer decisions.

If an organization meets every week but fewer and fewer people can actually call the shot, that's a warning. Everyone is syncing, updating status, explaining progress — but problems don't get closed. Meetings become a buffer against accountability: everyone participates, nobody truly owns the outcome. These companies look busy, but their ability to act keeps weakening.

Signal two: thicker reports, slower action.

Hollow companies love reports, because reports create the feeling that "things are being managed." But a report isn't a judgment. A status update isn't a decision. Progress narration isn't progress. If an organization gets better at explaining what it's doing — and worse at actually changing anything — it's hollowing out.

Signal three: AI is allowed to improve efficiency, not rebuild workflows.

This may be the most important signal. If you're allowed to use AI for emails, meeting notes, and slide decks — but not to question the process itself — the company hasn't really embraced AI. It just wants you to do the old work faster: it wants efficiency, not a rethink of whether the work should exist at all. Those are completely different things. Real AI transformation isn't automating old workflows. It's asking: Does this workflow still need to exist?

Signal four: people who understand the work lack power; people with power don't understand the work.

This is classic in large organizations. Frontline experts know where the problem is. People doing the work know how to fix it. But they can't make the call. People with authority don't always understand the work. So the organization becomes:

People who know, explaining.

People who don't, questioning.

People who can do, filling in materials.

People who can't judge, demanding more reports.

Real problems slow down. These companies are especially prone to hollowing out, because they exhaust capable people — until those people stop speaking up.

Signal five: the organization rewards storytelling, not problem-closing.

If the people who tell the best stories, build the best decks, and manage upward best get seen more than the people who actually solve problems, the organization is already tilting. Communication matters, of course. But when presentation outweighs real results, the system rewards performance — not creation. Problem Closers go quiet. Process actors multiply.

Signal six: high-agency employees go silent.

This may be the most dangerous signal. What companies should fear isn't complaining — complaints at least mean someone still cares. The real danger is when capable people stop talking: they stop suggesting, stop challenging processes, stop pushing for change. They do only what's assigned. They update their resume. They look outside. They pull their creativity back into themselves. When an organization silences its most agentic people, it's already losing the future.

Signal seven: the company can't answer "why are we irreplaceable in the AI era?"

This isn't just a personal question. Companies need an answer too. If AI can generate content, why do you still need this content team? If AI can write code, why do you still need this development process? If AI can organize tasks, why do you still need this project management system? If AI can generate reports, why do you still need this many reporting layers? If AI can connect user needs directly to execution systems, why do you still need this many middle steps?

A company that can't answer these questions won't necessarily die tomorrow. But it will get re-priced.

So how do you tell if a company is hollowing out? Not by whether it has money today, not by whether it has customers today, not by whether it's big today — ask instead:

• Is it still creating the future?
• Does it still attract capable people?
• Is it willing to cut old workflows?
• Does it allow AI to rebuild work?
• Does it reward Problem Closers?
• Does it give judgment to people who actually understand the work?
• Can it answer: in the AI era, why do we still matter?

This matters for ordinary people. Because choosing a company isn't just choosing salary, title, office, and benefits — you're choosing a system. If that system is hollowing out, the longer you stay, the more your capability gets compressed. You may get better at reporting — and worse at creating; more fluent in process — and further from the market; more secure — and less irreplaceable.

That's the danger. A hollow company doesn't destroy you all at once. It slowly drains you, letting you live in a still-decent environment while you lose sensitivity to the future.

That doesn't mean everyone should quit immediately, or resign the moment you spot a few signals. Real life has constraints — mortgages, family, visas, income, risk tolerance, career stage — and those are real. So the question isn't "Should I leave right now?" The real question is: If I see my company hollowing out, can I start building outside optionality for myself? Can I make my capability less fully dependent on this one organization? Can I start building work samples, content, relationships, product experiments, job pipelines, and personal systems?

I'll continue that in the next essay. Because in the AI era, the safest people may not be those with a stable job — they're the ones with options.

That's what Future Lab wants to keep studying. If legacy companies slowly hollow out, how should individuals read their environment? If AI re-prices companies, how should people re-price their own career choices? If an organization no longer amplifies your capability, how do you prepare your next move?

I don't have all the answers yet. But I'm increasingly sure of one thing: