TOP 6 Highest-Paying Careers for The Next 2+ Years (AI won't touch)

Executive summary

Nick Saraev’s thesis is that as AI agents take on more routine work, economic value will shift toward bottlenecks AI cannot yet remove: human judgment, physical work, accountability, trust-based selling, and ownership of productive assets. He recommends building skills around those constraints—and using AI to multiply output rather than competing with it.

Core frameworks and playbooks

• Bottleneck economics: Pay and opportunity accrue to the scarcest constraint in a process. As AI makes some tasks easier to produce, the valuable constraint shifts elsewhere—for example, from writing code to deploying it effectively inside a business.
• “Centaur” model: Let agents handle volume and execution; reserve human effort for judgment, direction, and accountability.
• Agent deployment workflow:
  1. Interview business owners and department leads to document processes.
  2. Map work into tasks agents can do independently and tasks that need human-agent collaboration.
  3. Capture workflows through recordings, screenshots, or transcripts, then turn them into standard operating procedures (SOPs).
  4. Automate suitable tasks and train employees to use agents on the rest.
  5. Check that results meet business needs and shift people toward higher-value work.
• RACE growth framework: Focus front-end business automation on Reach, Acquisition, Conversion, and Expansion.
• Ownership and leverage: Favor assets and reusable work—such as software, media, audiences, and customer relationships—that can generate returns beyond the hours spent creating them.
• Career-to-business path: Start by solving an urgent problem as a service provider, build expertise and reputation, then consider expanding into a business or scalable product.

Six careers and opportunities highlighted

1. AI agent operator / forward-deployed engineer

Works with a business to identify processes agents can handle, build the systems, and improve productivity. Saraev says smaller businesses often have more urgent needs in growth and customer-facing operations than in back-office automation.

Examples include automated email organization and AI-generated business assets or proposals.

Saraev cites U.S. median compensation of roughly $200,000–$250,000, around $400,000 for top earners, and an Anthropic posting around $300,000–$350,000. He also says Anthropic announced $100 million to train 10,000 such engineers by the end of 2027.

Action: Learn coding agents and process discovery. Focus on understanding the client’s workflow and delivering measurable business outcomes.

2. Electrician and related data-center trades

Data-center expansion creates demand for electricity infrastructure, installation, and maintenance—work the speaker considers difficult to automate in the near term.

Saraev cites data centers as consuming 4.4% of U.S. electricity in 2023 and guesses they could reach 8%–20% by 2028. He says electrical work accounts for about half of labor on data-center projects.

He estimates electrician pay at $65,000–$100,000 generally, versus $80,000–$140,000 for data-center installation and maintenance. He also recounts Mike Rowe’s anecdote about some Texas electricians under 30 earning $240,000–$280,000.

Pathways: Trade-school preparation of roughly 3–6 months, followed by a paid apprenticeship of 3–5 years toward licensing. Faster entry points include fiber-optic work, with certification taking days to weeks, or solar-installation training of 40–60 hours.

Action: Check local demand and consider relocating to areas with active data-center construction.

3. AI safety, red-teaming, and system evaluation

Tests AI systems for unsafe behavior, vulnerabilities, prompt injection, and unintended access to sensitive information. The business need rises as companies give agents access to data, money, and operational permissions.

Saraev cites an Anthropic red-team engineering salary posting around $400,000 and bug-bounty payouts of up to $35,000 for a vulnerability.

Action: Start by testing models responsibly and documenting reproducible issues. Security engineering experience and formal qualifications can help with more advanced roles.

4. AI media producer

Creates images, video, audio, or advertising assets with AI tools. As production becomes cheaper and faster, Saraev argues, taste and selection—choosing what audiences will value—become the differentiating skills.

He cites Upwork data showing demand for AI video generation rose 300% in 2025, calling it the platform’s fastest-growing skill.

For example, producers can generate locations and consistent actors instead of organizing an international shoot, or create voice work without waiting for studio availability.

Action: Learn a workflow across image, video, and audio tools; develop prompt styles and creative judgment; and use AI agents to generate and compare variations.

5. High-trust enterprise salesperson / corporate account manager

Closes large, complex deals where credibility, customer relationships, and in-person trust matter. AI can help with lead screening, qualification, outreach, and follow-up, but the salesperson remains central to relationship-building and negotiation.

Saraev cites an Anthropic corporate account manager range of $290,000–$380,000 and sales-engineer averages of about $125,000 in May 2025, which he estimates may have risen to around $150,000.

He argues that salespeople who generate $15 million–$20 million in annual revenue may earn at least $1 million in compensation; these are illustrative claims, not typical salary expectations.

Path: Begin in technical sales or customer support at a smaller SaaS company, then move toward larger, more expensive products and accounts.

Action: Build a record of sales performance. Use agents to support the funnel while prioritizing customer trust and deal execution.

6. Owner of an AI-leveraged business

The final opportunity is ownership rather than a specific job: own a company or asset in which a small team and AI agents generate much of the output.

Saraev cites Carta figures claiming solo founders accounted for 24% of new startups in 2019 and 36.3% in the first half of 2025.

As an example, he describes his own company as having himself and two full-time employees, with projected monthly revenue of nearly $800,000 at the time of recording. He says the business focuses on media and information products that can keep generating value after their initial creation.

Action: Don’t rush to build a product before validating demand. Start with a service addressing a current problem, build expertise and a customer base, then use that knowledge and distribution to develop scalable offerings.

Metrics and forecasts cited

• METR’s AI task “time horizon” reportedly doubled about every seven months as of March 2025. The speaker predicts continued growth and discusses the implications through 2028.
• Saraev says Anthropic found its agents spent six to seven times as many hours on tasks as all human employees combined.
• He cites the U.S. labor share at 54.1% in a first quarter, describing it as the lowest since measurement began in 1947.
• His central forecast is that automation will make execution cheaper, while judgment, physical responsibility, customer relationships, creative taste, and asset ownership become more valuable.

Presenter and sources referenced

• Presenter: Nick Saraev.
• Sources and examples mentioned: METR, Anthropic, OpenAI, the U.S. Department of Energy, Upwork, Carta, Mike Rowe, The Intelligence Explosion, and Pliny the Liberator.