New Research: 80-95% of Work Tasks Could Be AI-Automated by 2029

The largest worker-evaluated study of AI automation just dropped — and the findings challenge both the doomers and the deniers.

Researchers surveyed over 17,000 workers across 3,000+ job tasks derived from the U.S. Department of Labor's O*NET database. Their question: is AI automation hitting like "crashing waves" — sudden disruption in specific job categories — or "rising tides" — broad, gradual improvement across everything?

The answer: rising tides. And the water is rising faster than most people realize.

The Key Numbers

The study tracked AI's ability to complete real-world work tasks at minimum acceptable quality. Here's the trajectory:

• Q2 2024: AI successfully completed ~50% of text-based tasks requiring 3-4 human hours
• Q3 2025: That number climbed to ~65%
• Projected by 2029: 80-95% success rate across most text-related work tasks

Let that sink in. We're not talking about toy demos or cherry-picked benchmarks. These are real workers evaluating whether AI can do their actual job tasks to an acceptable standard. And the success rate is climbing roughly 10 percentage points per year.

Why "Rising Tides" Matters More Than "Crashing Waves"

Previous research suggested AI might automate certain jobs overnight — a wave that crashes into accounting, or legal research, or customer service, while leaving other professions untouched. That's the "crashing waves" model.

This study found little evidence for that pattern. Instead, AI is getting incrementally better at almost everything simultaneously. No single profession is being wiped out overnight. But no profession is standing still, either.

This is both reassuring and unsettling:

The good news: You probably won't wake up tomorrow to find your entire job automated. There's no sudden wave coming for your specific role.

The uncomfortable truth: The tide is rising under every desk in every office. The tasks AI couldn't do last year, it can do this year. The ones it can't do this year, it'll handle next year. And by 2029, the researchers project AI will be able to perform the vast majority of text-based work at acceptable quality.

What This Means for Your Career

1. The timeline is shorter than you think

Three years. That's the gap between today and a world where AI can handle 80-95% of text-related tasks. If you're planning to "wait and see" before adapting, you're already running out of runway.

2. No job is an island

The crashing waves model would let you relax if you were in a "safe" category. Rising tides means there's no safe category — just different water levels. A teacher, a marketer, a software developer, and a financial analyst are all seeing AI capability creep into their daily tasks, just at different rates.

3. The adoption gap is your window

Here's the critical nuance the headlines miss: the study notes that actual economic disruption lags technological capability by years. AI can do the task today, but most organizations won't restructure around that capability for another 2-5 years. That gap between "AI can do this" and "your employer expects AI to do this" is your adaptation window.

4. Augmentation beats resistance

The workers who thrive in a rising-tide world aren't the ones fighting the water. They're the ones who learn to swim — who use AI to amplify their uniquely human skills (judgment, creativity, relationships, leadership) while letting AI handle the tasks it's increasingly good at.

How Exposed Is Your Specific Role?

The study covered 3,000+ tasks, but your risk depends on your specific combination of work type, industry, routine level, and human interaction requirements. A marketing manager who writes ad copy faces different exposure than one who manages client relationships. A software developer building CRUD apps faces different exposure than one architecting distributed systems.

That's exactly what our AI career risk assessment measures. In 90 seconds, we score your specific job across 9 dimensions — including the factors this study highlights — using data from Anthropic, ILO, OECD, and BLS covering 800+ occupations.

Your score won't just tell you if you're at risk. It'll tell you which specific aspects of your role are most exposed and what to do about it — before the tide reaches your desk.

Take the free 90-second assessment →

FAQ

Is this study saying AI will replace 80-95% of jobs by 2029?

No. It says AI will be able to perform 80-95% of text-based tasks at minimum acceptable quality. That's different from replacing entire jobs. Most jobs involve a mix of text-based and non-text-based work, plus judgment, relationships, and physical presence that AI can't replicate. But if a large portion of your daily tasks become automatable, your role will change significantly — even if your job title stays the same.

Which jobs are most affected by the rising tide?

The study focused on text-based tasks assessable by large language models. Roles with heavy text processing — data analysis, report writing, customer communication, content creation, administrative coordination — are seeing the fastest capability gains. Roles requiring physical presence, real-time human judgment, or complex social interaction are less exposed, though still not immune.

What about the tasks AI still can't do?

Even at 80-95% task success, the remaining 5-20% tends to be the hardest, most valuable work: novel problem-solving, ethical judgment, stakeholder negotiation, creative strategy, and emotional intelligence. These are the skills worth investing in — they're the high ground as the tide rises.

Should I be worried?

You should be informed, not panicked. The study's own authors note that organizational adoption lags technology by years. But "years" is not "decades." The professionals who assess their specific exposure now, identify their vulnerable tasks, and start adapting have a meaningful advantage over those who wait.

---

This analysis is based on "Crashing Waves vs Rising Tides: Preliminary Findings on AI Automation from Thousands of Worker Evaluations of Labor Market Tasks" by Mertens et al. (2026). All projections cited are from the original research.

Want to know exactly where you stand? Take the free AI career risk assessment →