Relocating Human Value

Look closely at these companies and you get a clear preview of what the broader workforce is about to face: a world where AI handles the work that used to break in new employees, leaving little room for beginners to become experts, and where the gap between a skilled senior employee and no employee at all is narrowing every day.

The AI revolution is different from every technological wave before it. The internet created jobs. The smartphone created industries. But AI does not require a person at every step. It is the first massive technology wave whose core function is to reduce the need for human involvement, not channel it somewhere new.

This distinction matters enormously for a generation that was told: learn to code, get a computer science degree, and the jobs will follow.

Technology trends compound quickly. The internet, the PC, the iPhone, apps. With each new wave, the world reorganizes around it. The years 2023 through 2025 represent the most rapid periods of advancement in the history of generative AI, from emerging technology to a tool used widely across industries in under three years. By 2025, 92% of companies planned to increase their investment in generative AI over the following three years, according to McKinsey.

Just downstream of the companies building foundational AI models, OpenAI and Anthropic, are applied AI companies building products on top of those models and tailoring them to specific uses. This has created a new category: the AI-native startup. These are not companies that added AI to an existing product. They are companies where AI is the product, or at minimum, the skeleton everything else is built around.

MindTrip, a Silicon Valley-based AI travel planning startup, offers a clear example. Its chief marketing officer, Michelle Denogen, draws a useful distinction between two types of AI. Horizontal AI, like ChatGPT, can answer questions about healthcare, parenting, and car repair in the same conversation. Vertical AI goes deep into one domain. MindTrip is vertical, it focuses entirely on travel.

Going vertical meant building what Denogen calls a geo-tree — an indexed knowledge base of the world's travel destinations, pulling in pricing, reviews, photos and real-time data through third-party APIs, so that MindTrip's recommendations are more accurate and current than any general-purpose model could produce alone. But even this requires constant oversight. The model does not automatically recognize that a restaurant has closed or a hotel has changed ownership. Someone has to teach it that.

Then there is Plots, an AI-powered event discovery platform with over 400,000 monthly active users, founded by Tristan Amell. Amell started the company during a gap year in Boulder, Colorado, before attending UCLA, and he is not a trained engineer. Four years ago, when he first had the idea for Plots, he tried to build it using no-code app tools and quickly hit a wall. What has changed since then, he says, is everything.

Simon Shin, a Harvard grad, partner at Ethos Fund and one of Plots' investors, describes a clean arc where AI sits right now in relation to human work. It started as a co-pilot, assisting with productivity while still requiring a person to do the actual job. It has since become a collaborator, with AI agents performing entire job functions. The direction it is heading, he says, is toward operator, a state where organizations could theoretically be run entirely by AI, governed by a small human board.

For now, AI is not eliminating humans in the workplace. It is relocating them. But that relocation is not cost-free, and it is not evenly distributed.

Rocco Berbetti has been the team's sales lead for more than two years. In the past five months, his role has changed drastically. With no coding background, he built a tool, using AI, to handle his sales outreach. The tool identifies potential leads and does the initial contact for him, the task that used to occupy most of his day.

What that has freed up, he says, is time for the parts of his job he actually values: building genuine relationships with event organizers around the world. Before AI, he handled both the account executive role, closing deals, and the sales development role, finding leads and booking first meetings. Two distinct positions. Now the prospecting function is largely automated, and he has collapsed both into something more focused.

Tristan Amell, Plots' chief executive officer, is direct about what this means structurally.

It is not a hypothetical. A 2024 Bloomberg analysis found AI could replace 67% of sales representative tasks, exactly the prospecting and outreach work that SDRs traditionally perform. Rocco does not dispute that his role is changing. But he does not find it threatening.

Jan Prokes, Plots' front-end engineer, is based in Prague and joins the team's morning meetings remotely. As the person responsible for everything users see, for example, the follow button that turns from blue to grey and the feed that loads when you open the app, his work sits at the most visible layer of the product. And it has been transformed.

Prokes describes himself now as more of a supervisor than a coder. He tells AI what to build and steers it when it goes wrong, which it does, regularly. To manage that, he developed a system: he had AI generate a rule book, and every time it makes a mistake, he instructs it to log the error so it will not repeat it. The AI now has multiple indexed files it can reference when facing a specific engineering problem. He describes himself as the architect. AI is the builder. And the builder still needs substantial guidance.

Sang Lee, Plots' back-end engineer, tells a similar story. Like Prokes, he uses AI to write code and still acts as an overseer. Because AI handles more of the execution, he has taken on roles he would not typically be able to fill, functioning simultaneously as a project manager and an active developer.

The freedom AI brings is real. But so is a cost Lee has started to notice. His muscle memory for fine-grained details in code is quietly eroding.

It is a small signal of something larger.

Ed Francis is the chief technology officer and co-founder of GigPig, a startup based in Manchester, England, that built a digital marketplace connecting musicians with venues. He has watched the dynamic play out in real time, both inside his company and across the industry.

He calls what he sees in new graduates "vibe coding," prompting AI to generate code without possessing the underlying knowledge to evaluate what it produces. The output can look right and be wrong in ways that only experience would catch. And experience, increasingly, is what companies are requiring.

Francis is careful to say that GigPig does not use AI to replace staff. But he recognizes what is happening at the industry level.

The irony is that AI still needs human eyes. Prokes, Lee and Francis all agree: give AI a large, underspecified task and it produces bad results. Give it a strong foundation and let it fill in the details, and it works well. For now, the human role has not disappeared, its demands are simply different.

Lee frames the risk with a metaphor: if every room in a house is built by AI without a human architect setting the structural logic, the house will eventually crumble from within. Without supervision, AI builds a house of cards.

Denogen at MindTrip pushes back on the pessimism, at least about the technology itself.

She draws a comparison to Wikipedia. When it launched, teachers banned it and academics dismissed it. Now it is a default first stop for researchers, journalists and students worldwide. AI, she believes, is approaching that same tipping point, the moment when skepticism gives way to assumption.

But the pace of that shift is exactly what makes the entry-level problem urgent. The window between AI being good enough to take the job and good enough to be fully trusted is precisely the window where junior workers used to learn. And Anthropic CEO Dario Amodei has predicted that AI could eliminate as many as 50% of entry-level white-collar jobs within five years — a forecast that, even discounted, suggests the window is closing fast.

Tessa James, founder and chief executive of HireGains.ai, a Monaco-based startup that was founded before the AI boom and rebuilt itself as an AI-native recruiting company in 2025, sees the restructuring from a different angle. Her company uses a three-layer intelligence system to match candidates with employers: a 12-minute interview with an AI agent, a reflective evaluation for both hiring managers and candidates, and voice AI analysis to identify where candidates are most engaged.

Her primary use case is connecting domain experts in fields like financial services, law and healthcare with AI-native organizations that need that expertise to train their models.

Regulated industries, finance, healthcare, insurance, are being audited closely. AI can determine whether a mortgage gets approved or flag a potential cancer diagnosis. Getting those decisions wrong is not an inconvenience; it is a liability. So human-in-the-loop protocols are becoming requirements, not options, creating a category of roles for senior experts who can supervise AI in high-stakes environments. Those roles are safe, for now. Entry-level roles in the same fields are a different story.

But something else is happening alongside the contraction. A growing number of people, finding the traditional job market closed, are starting businesses on their own. Shin calls it the rise of the solopreneur, and the data backs him up. According to Carta's 2025 Solo Founders Report, the share of new startups with a single founder rose from 23.7% in 2019 to 36.3% by mid-2025.

The answer, he says, is that they probably would not. Kevin Terrell, founder of BirchAI, put it from the founder's side: "As a new startup, if I already have a few hundred thousand in revenue with a mix of customers, why would I give away 20% of my company for a $3 to $5 million investment?" These businesses are genuinely lucrative and meaningful for the person running them, but they are not built to scale toward an acquisition or an initial public offering. The venture capital model, Shin believes, may have to change because of them.

But the solopreneur path has a prerequisite: domain expertise. Building a marketing AI agent without knowing marketing produces a mediocre marketing agent. The barrier to starting a company has dropped. The barrier to starting a good one has not.

Amell is blunter about who this pressure falls on.

The one exception he carves out is creatives. At an events company, someone who can read a room and build a crowd brings something that does not translate into a prompt. And on the marketing side, he has noticed that the content performing best on platforms like TikTok is the content that feels most human.

He points to something a prominent tech executive said that stuck with him: that people who think differently, those with ADHD, autism, nonlinear minds, may actually have an edge in the AI age, precisely because AI is best directed by people who do not think in straight lines.

Shin agrees that new job categories will emerge, roles like agent orchestrator, titles that do not yet exist widely but are beginning to appear in job listings. He is insistent that the underlying skills of good thinking have not changed.

The AI revolution is not an apocalypse. It is a reorganization of which tasks get done by machines, of which human capabilities become more valuable, of what it means to be indispensable. The problem is not that work is disappearing. The problem is that the on-ramp is.