After Automation: AI progress creates more work for humans, not less

After Automation AI progress creates more work for humans, not less Dan ShipperCEO of Every There is a paradox at the heart of AI. At Every, we’ve automated everything we can. We use Codex and Claude Code across coding, writing, design, customer service, and more. We alpha-test all of the new models from OpenAI, Anthropic, and Google before they come out. We are riding the exponential boom in model intelligence and automation as far and as fast as possible. And yet it seems like, for us, there’s more human work to do than ever. We are a team of almost 30 people, and we haven’t fired all of our employees in favor of agents. We haven’t ditched software-as-a-service (SaaS) products in favor of vibe coded apps. We still hire humans to do customer service (with a lot of agent assistance), and we still hire human writers and editors and engineers. Our work does look completely different than it used to, though. We don’t write code by hand anymore. If you @-mention someone in our Slack, it’s a toss-up whether you’re talking to a human or an agent. Managers are committing code like ICs and engineers are talking directly to customers. For the last several weeks, AI has responded to 95 percent of my work emails. I am almost always at inbox zero (an extremely rare state for me), but I still review my email. In short, the future looks weird, but also familiar. The familiarity is surprising because the one thing CEOs, knowledge workers, and investors seem to agree on is that AI is a threat to jobs, the economy, safety, and human meaning. Anthropic CEO Dario Amodei warns that AI could wipe out up to half of all entry-level white-collar jobs(1). Meta just laid off 8,000 people, and is installing software on U.S. employees' computers to capture mouse movements, clicks, and keystrokes for a higher quality source of AI training data on advanced knowledge work. Even Citadel’s Ken Griffin seems shaken, saying recently, “These are not mid-tier white collar jobs. These are like extraordinarily high-skilled jobs being, I’m going to pick a word, automated by agentic AI.” The benchmarks, rising exponentially with each new model, seem to back this up. On Humanity’s Last Exam—a test of graduate-level reasoning—top models went from the low single digits a year ago to roughly 44 percent today. On GDPval—a test of how well frontier models perform real-world economic work versus humans—frontier models jumped from similar lows to about 85 percent. In May, METR, an AI safety research nonprofit, released an early Claude Mythos result showing that the model had an 80 percent success rate on tasks that would take a human expert about 4 hours to complete. It seems that we are on the cusp of an AI smarter than any human, with the autonomy to work for almost a full day at a time. And yet, the paradox remains. If you talk to anyone in the AI industry—or to early adopters outside of it—you’ll hear the same thing we’ve noticed internally: There’s more work to do than ever. The big question, within the industry and without, is: Is this just a temporary state of affairs? Will the next model drop be the one to replace everyone? We watch the benchmarks and sweat, wondering if there’s a tipping point around the corner where all of the jobs go away. There’s no tipping point coming where things flip and the jobs are gone. The new reality is the opposite—the more we automate, the more expert human work there is to do. Here’s why: AI commoditizes the residue of human expertise—whatever can be made explicit enough to train on. That collapses the value of default model output and creates demand for what’s different. Demand for what’s different is demand for human experts, even as we approach artificial general intelligence (AGI). To understand why this is, we have to go beyond the graphs, and look at how AI is used for work today. That will help us see, in a more grounded light, the paradox—and its resolution. 1 The three prompts from this essay, ready to run 2 The workflows we use at Every 3 Plus our guides and hands-on camp How we got here We’ve been covering the future of work with agents since 2022. Three years ago, I wrote about the allocation economy(2): that working with AI tools would eventually look a lot like the work of human managers. This was back when basic prompts and responses inside of ChatGPT were still considered alarmingly futuristic. Then, as a company, we became extremely Claude Code-pilled in mid-2025. Kieran Klaassen, general manager of Cora, suddenly found he was able to ditch hand-writing code in favor of spending all day giving plain-English instructions to a coding agent in his terminal. That quickly spread to the rest of the organization, and on Lenny’s Podcast 12 months ago I called Claude Code the most underrated tool for knowledge work. I bring this up because our best predictions have come from treating Every as a kind of early-adopter lab. We tend to run into new work patterns before they are normalized, and as the technology matures and the tools become easier to use, those patterns start showing up in the broader market. Here’s what’s happening internally now. The two modes of working with agents Work with AI is starting to settle into two very different modes(3). The first is the one the AI discourse predicted pretty well: agents as employees. These are agents you delegate work to. Some are agents that live in Slack, have names and jobs, and can be tagged when you want them to do something. Some are agents embedded in an ongoing workflow—like customer service—acting as always-on gatekeepers for repetitive tasks. The second mode is stranger and, in my experience, more important. It is human-agent collaboration in tools like Codex, Claude Code, and Claude Cowork. These are not just places where you hand off work. They are becoming operating systems for the work itself, where you and multiple agents use the same computer, at the same time, to do highly complex, original work that can’t be done by an asynchronous agent. In both of these modes you can use AI to automate and delegate much of your work—and both of these modes require you or another human in order to work well. Agent employees Agent employees are given a job, and go off to produce an answer, an action, a report, a draft, a triage decision, without you in the loop. These come in at least two flavors: coworker agents and embedded agents. Coworker agent A coworker agent is one you can, for example, tag in Slack, and ask to do work. It’s around whenever you need it. These are agents in the style of OpenClaw, or our in-house Plus One. Claudie Claudie is our consulting team’s coworker agent. Claudie writes sales proposals, creates the first draft of training decks, keeps track of the project todos, and more. Andy Andy is our editorial team’s coworker agent. She collects “nuggets”—good ideas for stories, pulled from our internal Slack—and turns them into digests and first-pass takes that our writers use to compile the daily newsletter. Viktor Viktor is a general-purpose agent who does work throughout the organization. We use him to gather growth metrics, analyze user surveys, and turn messy internal discussions into research memos and product recommendations. Embedded agents Embedded agents live inside a product workflow. They’re less flexible but can be very powerful for helping with repetitive tasks. Fin is the cleanest example—an agent embedded in our customer service platform that handles a lot of our support load through chat and email. In a recent week in May, Fin participated in 65 percent of 202 Every support conversations and closed 81 of them without a human, which is 40.1 percent of all actionable conversations. Embedded agents like this allow our customer service manager, Waqqas Mir, to spend less time responding to basic tickets, and more time building the system that responds to tickets and on complex cases that require high-touch interaction. Human and AI collaboration Across both forms—coworker and embedded—the pattern is the same. Employee agents take over more of the stable, repeatable, well-framed layer of work. But there is a lot of work that still requires a human being in the loop. We’ve found over and over that for any kind of complex task, the best way to get great work is to have an AI and a human going back and forth in the same workspace. This is what Codex, Claude Code, and Cowork are for. They allow you to spin up and delegate work to one or more agents across multiple chat threads. These agents have access to your computer and all of your sources of data. You can see each task the agent is doing and thinking about—and can interrupt at any time. And you’re responsible for managing the agents at the start and end of each one of their tasks, making sure it’s done well, and finding the next piece of work to do. Kieran calls this the human “sandwich”—we’re the bread on either end of the AI’s work. The human sandwich. Source: Every. The most obvious example is coding. The engineers at Every spend all day going back and forth with agents. They are planning new features or bug fixes, reviewing work that’s been done, and—if they use our compound engineering philosophy—tuning their system to get better over time. But this kind of collaboration goes way beyond coding. A new operating system for knowledge work Codex and Claude Code are becoming a new operating system for work. I spend nearly my whole day in Codex, running my SaaS tools through its in-app browser. It lets me bring my agent to everything I do—and operate at a level I couldn’t reach alone. Writing I composed this piece in Proof in the in-app browser of Codex. Codex watches what I’m writing and can spin up a subagent to do any task I need: writing the first draft of a paragraph, researching examples for the next section, copy editing. Writing this essay in Proof inside Codex. Source: Every. Email I do email the same way. Cora is my email client, and I run it inside Codex’s in-app browser—scrolling my inbox and talking through each item out loud with Monologue. Codex and Cora handle the rest. A Cora inbox sweep. Source: Every. Every agent needs a human You might already see, in the midst of all of this automation, where the humans come in. In every example, the agent needs a human in order for the work to, well, work. Someone has to point it at the right thing, decide whether the output is good, catch the places where it is wrong, and turn the result into a real-life decision or process. The further away an agent gets from a human who is in charge of making sure it works well, the less well it works. In our initial internal roll-out of agents as employees, we gave every employee an agent, but we soon moved back to agents that work for a particular team or the whole company rather than for individuals. Why? Agents need a lot of maintenance, and personal agents quickly get stale when the employees they were working with gave up on them. We have a team of AI engineers who are in charge of making sure our agents work well—and we’ll need them for the foreseeable future. Even something as simple as automatically building PowerPoint presentations can turn into a huge task. One of our PowerPoint automations includes 24 skills and 18 scripts and costs $62 in tokens to make a single deck. That is the first-order reason agents create more work for humans. But there is a second-order reason, too. Why automation makes more work for humans If you look at AI’s exponential trajectory over the last few years, and think about how its architecture works and where its powers come from, you’ll see clear feedback loops that create more human work. AI makes yesterday’s human competence cheap Current language models are trained on the visible residue of human competence(4): code, prose, images, support tickets, product specs, and more. They take all of it—the exhaust of s [truncated for AI cost control]