Last March, at a demo day held by a San Francisco startup accelerator, one team put up a slide. The title was short: "We don't use SaaS." The audience laughed, and before the laughter had even faded, a dozen investors raised their hands. That team walked away with the most business cards collected that day. The word "SaaSpocalypse"—a play on software's apocalypse—became the biggest buzzword in Silicon Valley in 2026, and the question it raises has now crossed the Pacific to land on the desks of Korean founders staring at their monthly subscription bills.
The Claim That Software Has Started Eating Itself
The logic behind SaaSpocalypse is simple. As AI agents begin automating complex workflows, a single agent can now take on tasks once split across separate subscription tools—CRM, project management, customer support, design drafts. If a company can have one AI agent organize its sales data and send follow-up emails automatically instead of maintaining an annual Salesforce contract, the logic that justified the existing subscription starts to wobble. The subscription software market was built on the premise that it makes tedious human tasks easy—but now agents are starting to handle those same tasks even more directly.
Several overlapping observations have fueled the debate. Since the second half of 2025, agent features from OpenAI, Anthropic, and Google have begun handling complex, multi-step tasks, and the latest Y Combinator batch has seen a noticeable rise in startups positioning themselves as replacements for specific SaaS categories. Stock prices of major SaaS players like Salesforce, ServiceNow, and Zendesk have repeatedly wobbled with each AI-agent-related announcement—a sign that investors are starting to take this structural shift seriously.
The Economist's decision to cover this trend head-on reflects the fact that the debate has spread beyond Silicon Valley's internal circles into mainstream business-strategy discussions. The prediction that "30% of the SaaS market will be replaced by agents by 2030" has gone from a radical scenario to something treated as a conservative estimate—and applied to a global SaaS market estimated at $4 trillion, that's not a number easy to dismiss. This isn't just a passing tech trend; it points to a shake-up in the software industry's revenue structure itself.
The shift feels close to home in the Korean context, too. A large share of domestic startups and solo founders run on a combination of Notion, Slack, Figma, HubSpot, and Zapier. A SaaS subscription stack running 150,000 to 500,000 won a month (roughly $110 to $360) isn't unusual. Add team collaboration tools, accounting software, and email marketing platforms, and some end up spending millions of won—several thousand dollars—a year. The claim that this stack could be replaced by an integrated agent within six months to a year sounds like pressure to rethink fixed costs altogether.
But SaaS Isn't Disappearing as Fast as It Seems
There's also considerable skepticism about this outlook. The claim that SaaS will be replaced by AI agents in the near term rests on a few hidden assumptions—and critics point out that not all of them hold up.
The most common counterargument is structural switching cost. SaaS tools don't just provide features. Notion holds years of accumulated document structures and team collaboration habits; Slack carries the history of channels with clients and partners; HubSpot has built up deal histories and sales pipeline data. Migrating all of that to an AI-agent-based system isn't simply swapping software. It involves data migration, team training, redesigning work practices, and adjusting how you collaborate with clients. That cost often far exceeds several months' worth of subscription fees—which is why the simple math of "cancel it and save" doesn't add up.
Existing SaaS companies are also fighting back hard. Salesforce has launched its own AI agent platform, Agentforce, and Notion has folded an AI assistant into its core feature set. HubSpot, too, has started bundling AI-driven automation into its existing subscriptions. These companies are restructuring themselves to absorb agents directly into their platforms. Right now, there's a race between how fast AI-native startups can take SaaS's place and how fast incumbent SaaS companies can absorb AI features. Which side wins that race is still an open question.
The actual maturity of these agents is another factor that can't be ignored. Current AI agents handle simple, repetitive tasks well, but still have real limits when it comes to reliably grasping complex business logic or organization-specific context. Declarations like "we run sales with agents, no Salesforce needed" get attention, but how many teams are actually running that reliably in production is a different question entirely. Most publicized cases are small pilots or applications to simplified workflows. Part of why the SaaSpocalypse narrative has spread so quickly is that it's mixed with more expectation and fear than actual operating experience. Rushing to replace a SaaS stack without accounting for this gap could leave a bigger operational problem waiting in the form of feature gaps and data disconnects.
What You Need Right Now Isn't a New Tool—It's a Task Inventory
So what does this debate actually demand of Korean solo founders and small business owners?
Job-automation research has confirmed a recurring pattern for years now: automation tends to replace specific "tasks" first, rather than eliminating entire jobs outright. Researchers in that field say the real impact of automation only becomes visible once you break a job down into its component tasks rather than treating it as one bundle. AI agents don't deviate much from this pattern. Notion doesn't get replaced wholesale all at once—specific tasks within Notion, like summarizing recurring meetings, drafting documents in a fixed format, or tagging and categorizing, move to agents first. This framing gives you a clear starting point for reassessing your SaaS stack: instead of asking whether you need to replace a tool entirely, ask first which tasks within that tool could move to an agent.
Here's the practical order of operations. List out the SaaS tools you currently use, and next to each one, write down the tasks you actually repeat on it. If tasks where "handing it to an AI agent wouldn't meaningfully change the outcome" account for more than half the time you spend on a given tool, that tool is a reasonable candidate for replacement in the medium term. But for tools where team collaboration history, customer data, and years of work records run deep, it's more realistic to wait for AI features to be built in natively than to switch right away. Before rolling out any new agent tool across the board, the right order is to pilot it on a single project for six weeks first, and only decide after directly confirming the switching cost and how feasible data migration actually is.
The specifics of the Korean market also need to be factored in. When collaborating with domestic companies, local tools like KakaoWork and Naver Works—Korea's homegrown answers to Slack—are often the de facto standard. Even if a global SaaS stack gets replaced by agents, domestic collaboration tools will likely need to stick around for a while for ecosystem reasons. The SaaSpocalypse debate is mostly aimed at the US enterprise market, and it starts from a different place than the criteria small and mid-sized Korean businesses use to choose tools. Before importing a debate spreading through the US wholesale, it's worth first checking the ecosystem of the market you actually operate in.
Here's what I'd say: what Korean founders need from this debate isn't a hasty decision to swap out their stack. It's an accurate picture of exactly which tasks their current tools are being used for, and how much. Swapping tools without a task inventory is like buying moving boxes before you know what you need to pack.
Teams that just sit back and watch SaaSpocalypse unfold will eventually pay the cost of switching late. Teams that rush to swap their stack out of fear pay an even more expensive tuition. Somewhere in between is the work that needs doing right now.
