In May 2025, a single line in a regulatory filing from ClickUp, the project-management software company, sent a tremor through Silicon Valley and beyond. The gist was this: rather than cut a workforce numbering in the hundreds, the company would deploy thousands of AI agents. It wasn’t an earnings announcement. It wasn’t a product launch. It was a personnel decision.
What made the document arresting wasn’t the number itself but the context the number sat in. Until now, the forecast that “AI will replace jobs” had been future-tense talk for the conference stage. ClickUp’s filing took that forecast and reclassified it as an operational decision from last quarter. It happened not in a keynote speaker’s slide deck but in the kind of formal corporate document a company files with the SEC.
This column starts from that number. It sets out to ask what the number means for Korea’s solo entrepreneurs, solo PMs, and content directors.
ClickUp’s filing, and two more signals from the same stretch
ClickUp’s announcement wasn’t an ordinary restructuring notice. The company explicitly named AI agents as the means of replacing the eliminated positions. It’s the first case in the SaaS industry of an agent being pinned down, in a public document, as a “cost-cutting tool.” Efficiency through automation existed before, of course. But few companies had ever declared so openly that they were swapping out the headcount of a specific job category for a number of agents.
Around the same time, there were two more signals.
One came from academia. A paper called EVE-Agent was released, and the research showed in concrete terms just how fragile AI agents can be in real-world settings. Agents are being deployed fast, the paper argued, but the trust architecture that governs which decisions an agent makes, and in what context, is still unfinished.
The other came from the Vatican. The Holy See formally issued an encyclical on the subject of AI. A papal encyclical is the most authoritative document the Catholic Church issues on matters of faith and morals. That such a document addressed AI governance directly is a signal that the topic has spread beyond the tech community’s internal debates and into global institutions as a whole.
These three events belong to different domains, but they point in the same direction: the gap between how fast agents are being deployed and the norms meant to govern them is widening quickly.
Why this round of agent replacement is different
The line “AI threatens jobs” has been repeated since 2016. When an Oxford University study announced that “47% of US jobs are at risk of automation within the next two decades,” and when McKinsey warned that “up to 800 million jobs worldwide could be displaced by 2030,” many people received the prediction in the future tense. It read like a story that didn’t quite touch them directly.
What sets ClickUp’s filing apart is that it was written in the past tense. Not “we’ll look into this” but “we’ve already decided.” Hundreds of seats emptied, and agents are being placed in those empty seats. This is a far more direct market signal than model-performance benchmarks or GPU-race metrics.
Here’s the part worth noting: the work ClickUp is moving to replace isn’t only simple, repetitive tasks. Customer support, content operations, internal-communications handling — domains once thought to “require human judgment” — are being named as targets for agent placement. These are exactly the domains that Korea’s solo entrepreneurs and solo operators have handled alone, outsourced to freelancers, or run with a small team.
A large share of the work a Korean solo entrepreneur outsources to freelancers — or handles personally — falls into this category. Work that is repetitive, that can be structured, and whose output quality can be measured in numbers.
Add one perspective often raised in the context of people management, and the picture gains depth. The role organizations expect from their members is shifting fast — from “someone who executes a given procedure precisely” to “someone who reads context and makes judgment calls.” That shift seemed like it would take its time, but ClickUp’s filing shows it has already advanced to the level of a personnel decision.
At the same time, the problem the EVE-Agent paper points to — that the trust architecture for agents is unfinished — matters just as much. Agents aren’t flawless. They make judgments in unforeseen ways and act in the wrong context. Which means you need a person to oversee, review, and course-correct the work agents handle. That role doesn’t get automated.
The papal encyclical’s mention of AI governance lives on the same plane. Technology is being deployed so fast that the question of who is responsible, and how, for the outcomes it produces has risen to become a challenge for institutions across the board. The Vatican document, too, carries the judgment that industry self-regulation alone can’t keep pace.
What solo operators should audit right now
The first question a Korean solo entrepreneur can ask after reading ClickUp’s filing is this: “Of the work I do right now, how much of it could an agent handle?”
The reason the question is uncomfortable is that the answer may be larger than expected. And facing that discomfort is, at this moment, a practically useful place to start.
A few items are worth auditing.
Check for repeating structure. Make a list of the work you handle in a similar format every week or every month — drafting reports, email-response patterns, scheduling social content, sorting invoices. If an agent can take any of these to roughly 80%, you can hand that work to the agent and reassign your own role toward reviewing the remaining 20%.
Identify the points of judgment. Write down, fresh, why your clients pay you. Whether it’s fast delivery, judgment tuned to a particular taste and context, or trust built through the relationship — the range an agent can replace shifts accordingly. The more judgment and relationship sit at the center, the more the agent becomes a tool and you become the one accountable for the result.
Prioritize agent experiments. You don’t need to automate everything right now. But it’s worth picking one or two tasks and actually using agent tools yourself. When you apply tools like ChatGPT, Claude, Notion AI, and Make.com to your own real work, it becomes clear where the agent gets stuck and where your intervention is required. Talking about agents without that experience is a bit like reading swimming theory without ever getting in the water.
Recognize the governance risk. The trust-architecture problem the EVE-Agent paper raised applies directly to solo operators too. If you pass an agent’s output to a client without review, you bear the risk. Unless you set the scope of agent use and the review standard in advance, speed goes up but trust gets chipped away.
As for tool combinations worth experimenting with in the current market — it varies by type of work, but — for document work, consider Claude or ChatGPT; for automating repetitive processes, Make.com or Zapier; and for project tracking, Notion AI’s integration features. The goal isn’t to use these tools for their own sake; it’s to see for yourself, in your own work, the range an agent can actually handle.
ClickUp’s decision to replace hundreds of workers with agents was aimed at the company’s shareholders and board, but the ripple reaches the work structure of solo entrepreneurs too. If this is the pace at which agent deployment is being formalized as a personnel decision at the company level, then the nature of the work freelancers and solo operators are commissioned to do changes as well. Demand for pure execution work shrinks, and demand for the role of judging and adjusting an agent’s output grows.
How you read what has already happened, and how you fold it into your own way of working, becomes the fork in the road within the next year. ClickUp’s filing showed — in the names of hundreds of people — that the fork is closer than you’d think.
