AI and Jobs: What Three Decades of Building Tech Taught Me About What's Coming

In 2023, I started warning people. Friends. Family. Anyone who would listen.

I told them AI would upend their careers within three years. Most nodded politely and moved on. Some laughed. A few got defensive. Almost nobody took it seriously.

It’s 2026 now. I was right. I wish I hadn’t been.

Who Am I to Say This?

I’ve spent thirty years building what’s next before most people knew it was coming.

My earliest partner was Craig Newmark. We co-founded DigitalThreads in San Francisco in the mid-90s — Craig credits me with naming Craigslist and the initial setup. That project reshaped classified advertising, employment, and real estate. We didn’t know it would. We just knew something was shifting.

I founded Buzzmedia in 2004, backed by Redpoint Ventures, NEA, and Sutter Hill — before YouTube monetization, before Instagram, before “influencer” meant anything. We built over 30 interconnected communities that became top-100 internet destinations for music and celebrity content, reaching 110 million monthly users at peak. We discovered Kim Kardashian through community interest in her photos before she was famous. We gave platforms to Fall Out Boy, Paramore, and Adele fans before anyone thought fan communities mattered.

I co-founded Metapa with Scott Yara, backed by Ed Sim at BoldStart Ventures. When the dot-com bubble burst, we pivoted from content distribution to big data. That company became Greenplum, which EMC acquired.

I’ve built SaaS platforms and web communities through multiple technology cycles. I’ve watched the internet reshape media, mobile reshape computing, social reshape communication, streaming reshape entertainment.

I think in decades. That’s how I see what’s coming.

What’s coming is bigger than any of those shifts. Combined.

The Three-Year Warning Is Over

Three years have passed. Look at what’s happened.

AI models aren’t research projects anymore. They’re production systems inside the world’s largest enterprises. Anthropic went from zero to 44% enterprise market share in under two years. Claude Code hit $1 billion in run rate revenue six months after launch. Uber deployed it wall-to-wall across engineering, data science, finance, trust and safety. Salesforce rolled it to their entire global engineering organization. Accenture has tens of thousands of developers on it.

The capability I saw emerging in 2023 is now shipping code, writing contracts, analyzing financials, generating research. At scale. In production.

Dario Amodei, Anthropic’s CEO, said publicly that AI could eliminate half of all entry-level white-collar jobs within five years, potentially pushing U.S. unemployment to 10-20%. He called it a “white-collar bloodbath.”

Ford CEO Jim Farley warned AI “will replace literally half of all white-collar workers.” Salesforce’s Marc Benioff said AI already handles half the company’s workload. Kai-Fu Lee validated the projection: 50% of jobs displaced by 2027.

AI was cited in nearly 55,000 U.S. layoffs in 2025. Positions requiring 0-2 years of experience are vanishing at three times the rate of mid-career roles.

My three-year warning wasn’t pessimistic. It was accurate.

I was trying to be nice. I softened the message because I didn’t want to sound negative amid the hype. I hedged. I qualified. I gave people room to dismiss what I was saying.

I’m done with that. People need to understand what AI at industrial scale actually looks like — not the polished demos, not the productivity tips. The restructuring of how work gets done and who gets to do it.

How I Think About Technology Change

I think in decades, not years. Big technologies don’t arrive and change everything overnight. They arrive, get dismissed, get adopted slowly — then suddenly they’re everywhere.

The pattern repeats.

When ATMs appeared in 1967, people assumed they’d eliminate bank tellers. They didn’t — not immediately. According to the American Enterprise Institute, ATMs reduced the cost of running a branch, which led banks to open more branches, which meant teller employment kept growing until 2007. Three decades. Electricity took thirty years to reshape manufacturing after the first power plants fired up. The adoption curve is always longer than people expect.

But AI is different. The capability curve is exponential. The adoption curve remains logarithmic. The gap between what’s possible and what’s deployed is where we live right now — and that gap is creating a false sense of security.

Harry DeMott, partner and managing director at CO/AI, wrote about this in The Species That Wasn’t Ready. Ninety percent of American businesses still don’t use AI in production. Enterprise adoption crawled from 3.7% to 9.7% over two years, according to Anthropic’s research with Census Bureau data. Two years of the fastest capability improvement in computing history, and fewer than one in ten businesses deployed it.

That’s the “gradually” phase.

The “suddenly” phase comes next.

The Next Three Years: 2026-2029

Between now and 2029, product teams will build AI systems that transform every major industry using computing as a central resource. Not incrementally. Completely.

Legal: AI already reads contracts, summarizes case law, drafts briefs, and conducts legal research at a level that rivals junior associates. Within three years, large portions of document review, contract drafting, and legal research will be automated. Firms that adopt will operate with a fraction of the associates they employ today.

Accounting: AI handles financial modeling, data analysis, report generation, audit preparation. Work that takes teams of accountants weeks will take hours. Firms that adopt will need far fewer staff. Firms that don’t will lose clients to those that did.

Banking and Finance: Investment memos, credit analysis, risk assessment, compliance monitoring — all automatable. Goldman Sachs research identified these as the highest-risk occupations for displacement. And Goldman isn’t just warning about it — they deployed Claude six months ago with embedded Anthropic engineers building autonomous agents for accounting, compliance, and client onboarding. The banks writing the research reports are already automating.

Software Development: This is the field I know best. A year ago, AI could barely write a few lines of code without errors. Now it writes hundreds of thousands of lines that work. Claude Code hit $1 billion in revenue six months after launch. The head of Claude Code, Boris Cherny, hasn’t written any code himself in over two months. At Anthropic, AI writes 70-90% of all code company-wide. Four percent of GitHub public commits are Claude Code right now; analysts project 20% by year end. Vibe coding tools like Lovable and Cursor have seen combined valuations jump from $8 billion to $36 billion in eighteen months. Anthropic now hires generalists instead of specialists — traditional programming skills matter less when AI handles implementation. There will be far fewer programming jobs in a few years.

Medical Research and Back Office: AI reads scans, analyzes lab results, suggests diagnoses, reviews literature. Healthcare administration — billing, scheduling, records — is being automated rapidly.

Entertainment and Content Creation: This is another area I know deeply. Entertainment will be completely retooled. The stories stay the same — humans crave narrative, emotion, connection — but how those stories get made will change entirely. ByteDance’s Seedance 2.0 generates multi-scene narratives with consistent characters, quad-modal input, and native audio sync. Google’s Veo 3.1 creates 4K video with dialogue, sound effects, and ambient noise generated natively. Nano Banana Pro turns photos into production-ready imagery in one click. These are early indicators of what every content creator will have. What’s remarkable in 2026 will be shocking by 2031. Pre-production, storyboarding, visual effects, editing, sound design, music composition — every phase of the pipeline is being transformed. Gaming, print, audio, music, movies, television. The creative industries that shaped culture for a century are about to operate with a fraction of the workforce.

By 2029, the question won’t be whether AI can do these jobs. It’ll be whether humans can compete with AI’s cost and speed.

2029-2031: The Completion of the Shift

This is where decade-thinking comes in.

By 2031 — nearly a decade after I started warning friends and family, nearly a decade after we started publishing about how to prepare — the power shift from human-powered computing to AI-powered computing will be largely complete.

Significant corporate jobs will be gone or unnecessary.

Think about how massive data centers operate today. A few dozen cloud administrators run infrastructure supporting millions of companies’ products. The ratio of humans to output is radically different than twenty years ago. That’s what’s coming to white-collar work.

Yes, the technology is exciting. Yes, it feels magical. But it allows corporations to change how they operate, hire, and manage. Capital markets favor efficiency and profit. AI is changing both.

By 2031, corporations will not require half the workforce they do today. Even with market growth, the headcount won’t be needed. AI tooling, compute resources, and skilled AI talent will be so efficient that hiring additional people becomes a drag on performance.

The employees building rogue AI agents in their companies right now — the ones Rick Grinnell documented in his survey of 50+ CISOs — aren’t malicious actors. They’re early adopters proving one person with AI tools can do what used to require a team. When management catches on and institutionalizes that capability, the team becomes unnecessary.

The One-Person Billion-Dollar Company

By 2031, we’ll see companies with headcounts of 1 to 15 people generating revenues in the billions.

This isn’t speculation. Sam Altman said publicly he’s part of “a little group chat with my tech CEO friends” betting on when the first one-person billion-dollar company happens. When asked for his prediction, Dario Amodei answered: “2026.”

The early signs are already here. Cursor hit $500 million ARR with fewer than 50 employees — a five-fold increase since January. Lovable reached $17 million ARR with 15 employees, three months after launch. Gumloop raised $17 million with two full-time staff and wants to reach a billion-dollar valuation with ten.

These companies may never need to hire. May never need to go public. Why would they? The cash flow is extraordinary. EBITDA through the roof. Going public is a hassle. Just run it and take the money.

The Data Exchange Economy

Imagine a company run by one person trading clean, structured data to 100 million AI agents that need it like oxygen. The agents require that data to function. One company has it.

This isn’t fantasy. Microsoft researchers describe an “open agentic economy” — a web of agents forming a decentralized ecosystem where AI agents interact freely. PubMatic launched AgenticOS in January 2026, an operating system for agent-to-agent execution. Protege raised $30 million to build a governed marketplace for AI training data. Cloudflare acquired Human Native, an AI data marketplace.

It will be cheaper for AI agents to buy data in a bid-trade system than to build it themselves. Like ad exchanges on Google and Facebook now — fully AI-driven programmatic buying. But with one person running the operation, revenue in the hundreds of millions, headcount of one, computing costs as the only real expense.

Why This Changes Capital Markets

AI at scale doesn’t just change companies. It changes what capital markets are built on.

The traditional model assumes companies need to hire to grow, need to go public for capital, need large teams to generate large revenues. AI breaks all of it. When one person can generate the revenue of a thousand-person company, what does “market cap” mean? When the most profitable companies have headcounts you can count on two hands, what does “job creation” mean as an economic indicator?

As one analysis put it: “The founder of the future isn’t the one who raises $100 million and hires a hundred people. It’s the one who runs $10 million in revenue with five.”

The Economic Model Nobody’s Built

I’m not an economist. I want to be upfront about that. I read the data I could find — Goldman Sachs, the CBO, the Richmond Fed, the World Economic Forum — and I used AI to analyze it, stress-test my assumptions, look for holes. What follows isn’t academic modeling. It’s pattern recognition from someone who’s watched technology reshape industries for thirty years.

What troubles me: the economies worldwide haven’t priced this in.

I’m not aware of any serious economic model that accounts for 50% productivity increases at modern companies while headcount decreases at scale. The Congressional Budget Office, Goldman Sachs, the World Economic Forum — they’re all modeling productivity gains, but most assume displacement is temporary or offset by new job creation.

That assumption has historical precedent. The Richmond Fed notes that 60% of U.S. workers today are in occupations that didn’t exist in 1940. Technology has historically created more jobs than it destroyed.

But AI is different. It’s not replacing one skill. It’s a general substitute for cognitive work. It improves at everything simultaneously. When factories automated, displaced workers retrained as office workers. When the internet disrupted retail, workers moved into logistics or services. AI doesn’t leave a convenient gap. Whatever you retrain for, AI is improving at that too.

Dario Amodei acknowledged this: “Most of them are unaware that this is about to happen. It sounds crazy and people just don’t believe it.”

What happens when corporations operate at 2x efficiency with half the workforce? When that becomes the competitive standard? When companies that don’t adopt get outcompeted by those that do?

The economists haven’t modeled it. The politicians haven’t legislated for it. The workforce hasn’t prepared for it.

This Is Troubling. Perhaps a Disaster. But It’s Happening.

I’m not celebrating this.

Displacing tens of millions of workers is a potential catastrophe. The social fabric depends on people having meaningful work and economic participation. If corporations achieve the same output with half the humans, and those humans have nowhere to go because AI improves at every alternative simultaneously, we have a problem no economic system is designed to handle.

But whether I celebrate it or mourn it doesn’t change that it’s happening. The capability exists. Deployment is accelerating. Capital markets reward efficiency. The technology improves faster than institutions adapt.

Three Turing Award winners — Geoffrey Hinton, Yoshua Bengio, and Yann LeCun — are warning about the same timeline. Safety leads are leaving Anthropic and OpenAI. The people closest to the technology are sounding alarms.

Meanwhile, Anthropic raised $10 billion at a $350 billion valuation. The builders are walking away warning about what they built. The building isn’t stopping.

What You Should Do

Preparing people matters more than anything else I could be doing.

Use AI seriously. Not as a search engine — as a work partner. Sign up for Claude or ChatGPT’s paid version. Use the best model available. Push it into your actual work. If you’re a lawyer, give it contracts. In finance, give it your data. A manager, use it to analyze your team’s performance. The people getting ahead aren’t using AI casually. They’re actively automating the parts of their jobs that used to take hours.

Spend an hour a day experimenting. Not reading about AI. Using it. Every day, try to get it to do something new — something you haven’t tried, something you’re not sure it can handle. Do this for six months and you’ll understand what’s coming better than 99% of the people around you.

Get your financial house in order. Build savings if you can. Be cautious about debt that assumes your current income is guaranteed. Consider whether your fixed expenses give you flexibility or lock you in.

Lean into what’s hardest to replace. Relationships built over years. Work requiring physical presence. Roles with licensed accountability. Industries with regulatory hurdles. None are permanent shields. But they buy time. And time is the most valuable thing you can have — if you use it to adapt.

Rethink what you’re telling your kids. The standard playbook — good grades, college, stable professional job — points directly at the most exposed roles. If you have kids in grade school, understand that by the time they graduate high school, the way employment works will have changed entirely. If you have kids in high school, push the schools to teach AI literacy — it will be the skill they need to get a job. If you have kids in college, make them use AI daily and learn to build things themselves. They have to be generalists, comfortable thinking with their hands as much as their minds. Nobody knows what the job market looks like in ten years. The people most likely to thrive are the curious, the adaptable, the ones effective at using AI to do things they actually care about.

Final Thought

I told my friends and family three years ago that AI would challenge their careers. Most didn’t listen.

I’m telling you now: the next five years will determine whether you navigate this transition or get disrupted by it. The capability curve is exponential. The deployment curve is catching up. The gap between them is closing fast.

The people who thrive will be the ones who engage now. Who learn the tools. Who adapt their skills. Who build the muscle of continuous learning.

The people who wait will find themselves competing against AI — and against the humans who learned to work alongside it.

We used to tell our kids to go to college to prepare. I’m not sure that’s logical anymore.

What’s logical is learning to work with the technology reshaping everything. Building financial resilience. Being honest about what’s coming rather than hoping it doesn’t arrive.

The future is already here. It just hasn’t knocked on your door yet.

It will.

Join the Conversation

The one-on-one warnings weren’t enough. I kept having the same conversation, and it couldn’t scale. A group of us who saw the same thing coming decided to do something about it. CO/AI became our way to reach people at scale — thinkers, engineers, creatives, college students. Everyone who wants to understand what’s coming and prepare for it.

Anthony Batt is a hands-on entrepreneur who has founded several venture-backed startups over three decades, including being credited with naming Craigslist with Craig Newmark. Currently working with stealth AI startups. Member of CO/AI, a community focused on AI literacy for creative and business professionals. Follow CO/AI’s newsletter Signal/Noise for weekly analysis of how AI is reshaping work and industry.

AI and Jobs: What Three Decades of Building Tech Taught Me About What’s Coming