Reading Time: 10 minutes

Hey Prompt Lover,

Welcome to June.

Let me tell you something that would have sounded made up in 2022.

There are people right now being paid $130,000 to $220,000 a year to try to trick AI into saying something dangerous. Their job title is AI Red Teamer. It did not exist as a real profession three years ago. Today Microsoft has an entire team dedicated to it, Anthropic hires for it constantly, and entry-level postings are sitting at $60,000 to $70,000 for people who have never done it professionally before.

That is one role in a category of work that has appeared so fast most people haven't caught up with it yet.

The conversation about AI and jobs almost always goes in one direction. Which jobs is AI taking. Which professions are at risk. Which skills are becoming obsolete.

That conversation is real and worth having. But there is a parallel conversation that almost nobody is having.

What jobs did AI create that didn't exist before. What do they actually involve. What do they pay. And are they real careers or just a moment in time that pays well until the moment passes.

Today we go through them honestly.

700+ teams have Viktor reading their Google Ads every morning.

Your media team opens Slack at 8am. There's a cross-platform brief in #growth: Google Ads spend vs. ROAS, Meta CPA by campaign, Stripe revenue by channel. Viktor posted it at 6am. Nobody asked for it.

Last week, one team's Viktor caught a spend spike at 2am on a broad match campaign and flagged it in Slack: "CPA up 340%. Recommend pausing and shifting budget to the top two performers." That would have burned $3K by morning. The media buyer woke up to a problem already handled.

Your strategist reviews spend trends. Your account manager checks revenue attribution. Same Slack channel, same colleague, before anyone's first coffee.

Google Ads, Meta, Stripe. One message. No Looker, no Data Studio. Anomaly detection runs around the clock. Cross-platform reporting runs on autopilot.

5,700+ teams. SOC 2 certified. Your data never trains models.

"Viktor is now an integral team member, and after weeks of use we still feel we haven't uncovered the full potential." — Patrick O'Doherty, Director, Yarra Web

Start free. $100 in credits →

This is the one that went mainstream first and also the one that gets the most confused coverage.

Prompt engineers in the United States earn a base of $95,000 to $206,000 in 2026, with a national average near $129,500 across public salary aggregators. Frontier lab packages at Anthropic and OpenAI push total compensation past $500,000 once equity and signing money are included.

But the title covers an enormous range of work. At the low end it is someone who writes instructions for a chatbot. At the high end it is someone designing complex AI systems, building retrieval-augmented generation pipelines, running evaluation frameworks, and consulting on AI strategy at an enterprise level.

Entry level freelance sits at $50 to $80 per hour for simple prompt writing and chatbot setup. Mid-level runs $100 to $150 per hour for system prompt design and RAG implementations. Senior specialists command $150 to $250 per hour for AI strategy consulting, production system architecture, and team training.

The durability question is the one people ask most and the honest answer is that the title may not survive but the skill will. The people doing this work at the highest level are not just writing prompts. They understand how models behave, how to evaluate outputs systematically, and how to build AI into production systems. That knowledge compounds. The job title may evolve. The value of knowing how these tools actually work will not disappear.

How to get started. Build things. Document what you build. The market rewards demonstrated output over credentials. A portfolio of real prompt systems you have built and can show is worth more than any certification.

This one surprised more people than any other on this list.

RLHF stands for Reinforcement Learning from Human Feedback. It is the technique that turned raw language models into tools people actually want to use. The way it works is that humans evaluate pairs of AI responses and rank which is better, safer, or more helpful. The model learns from those rankings. Without this process most AI tools would be significantly less useful than they currently are.

The role exploded fastest after ChatGPT launched. RLHF reviewers evaluate AI-generated responses to help models learn which outputs are helpful, accurate, and safe. Entry points start at $14 to $22 per hour on platforms like Surge AI and DataAnnotation.

But the ceiling is dramatically higher than that entry point suggests.

Pay ranges span $15 per hour for entry-level annotators to $1,000 per hour for C-suite domain experts, with most mid-tier RLHF and AI tutor work falling in the $50 to $100 per hour range. Mercor, which went from roughly $1 million to $500 million in revenue in 17 months, screens contractors for expertise and pays an average of $95 per hour.

Specialized domain trainers in coding, math, and science earn $40 to $80 per hour or $80,000 to $120,000 full-time. Senior RLHF specialists at major labs earn $120,000 to $180,000 or more full-time with benefits.

The key insight here is that domain expertise is the multiplier. A general evaluator earns general rates. A doctor, lawyer, or senior developer evaluating outputs in their field earns significantly more because the model needs that specific expertise to improve.

The single most effective way to increase your AI trainer salary is domain specialisation. A general RLHF trainer earning $30 per hour who adds medical expertise can immediately access $100 plus per hour projects on the same platform.

How to get started. Sign up for Outlier AI, DataAnnotation, or Mercor. Start with general evaluation work to understand the methodology. Then position your existing professional expertise as a specialisation.

This is the role most people have never heard of and it is one of the most interesting on the list.

The AI Red Teamer proactively tests AI systems, especially LLMs and generative AI, for security vulnerabilities, safety risks, biases, and failure modes through adversarial simulation. Microsoft formed the first dedicated AI Red Team in 2018 but the field exploded after 2023 with the rise of LLMs. They have red-teamed over 100 generative AI products and published a whitepaper on their methodology.

The job is essentially trying to break AI systems before bad actors do. Finding the prompts that make a model produce dangerous content. Testing for bias and discrimination in outputs. Identifying ways proprietary information could leak. Looking for failure modes that would embarrass the company in public or create legal liability.

AI Red Teaming demand grew by 200% in 2024. Positions command six-figure salaries with a median total pay near $178,000, with some postings reaching well into the $200,000 range.

The full range sits at $130,000 to $220,000 with a 55% growth rate.

What makes this role genuinely accessible is that it values thinking style over traditional credentials. The field values demonstrated skills over formal certifications. Mercor lists psychology, acting, or writing backgrounds among desirable qualifications for unconventional adversarial thinking. Industries hiring include Microsoft, Google, Nvidia, OpenAI, AI security startups, defense contractors, financial services, and consulting firms.

How to get started. Learn the OWASP LLM Top 10 vulnerabilities. Practice red teaming open-source models. Build a portfolio of documented attempts and findings. The community around this is active and visible on X and GitHub.

Most AI tools forget you the moment the chat ends. SureThing doesn’t.

SureThing is an autonomous agent that can draft in your voice, triage what matters, follow up on things you forgot, and report back with what happened next.

Day 1, you onboard it.
Day 30, it knows your clients and patterns.
Day 90, it catches things you missed.

Try SureThing

This one is less visible than the others but it is where a lot of money is moving right now.

McKinsey's 2025 surveys show companies are still struggling to translate AI pilots into business results, and that gap is creating a market for new roles that go well beyond developers. Think ethics experts who prevent companies from making headlines for the wrong reasons, creative directors who can wrangle AI tools without losing brand voice, and strategists who know when to say maybe we shouldn't automate this just because we can.

The AI integration consultant sits in the gap between what AI can technically do and what a business can actually deploy. They figure out which processes to automate, how to do it without creating new problems, how to get a non-technical workforce to use these tools properly, and how to measure whether any of it is working.

89% of companies said AI was creating new roles including MLOps engineer and AI architect, although most also worry about retention and ROI. 78% of IT job roles now list AI expertise as a requirement.

This role is still being defined which means the people doing it well right now are building something that is genuinely theirs. Senior consultants are earning $150,000 to $300,000 depending on industry, with healthcare, finance, and legal commanding the highest rates.

How to get started. Pick one industry you already understand. Learn the AI tools that apply to it specifically. Find one company in that industry struggling to implement AI and help them for less than your eventual rate. Build the case study. That one case study is your entry point to everything else.

This is the honest question and the answer is that it depends on where in each category you land.

The low end of every role on this list is vulnerable. Basic prompt writing. Simple annotation. Entry-level testing. These tasks are exactly the kind that AI itself will get better at handling over time. The people building careers in that band are working against a clock.

The high end of every role is different. The people who combine deep domain expertise with real understanding of how these models work — the RLHF specialist who is also a practicing doctor, the red teamer who genuinely thinks like an adversary, the integration consultant who understands both the technology and the business problem it's supposed to solve — those people are building something that compounds.

AI engineers are experiencing exceptional demand with average salaries soaring to $206,000, a huge $50,000 increase from the previous year. 97 million new roles may emerge by 2025, creating a net positive of 12 million jobs globally. But only for those who understand that AI isn't eliminating work — it's reshaping where value gets created. These jobs aren't theoretical. They're being posted right now, with real salary ranges, by real companies with real budgets.

The through line across every durable role on this list is the same. Specific expertise plus real understanding of how AI actually works. Neither one alone is enough. Both together is the combination that compounds.

Start with the role that maps most naturally to what you already know. If you have domain expertise in a professional field, RLHF training and AI evaluation work pays you immediately for knowledge you already have. If you think like an adversary and enjoy finding holes in systems, red teaming has the clearest entry path with the lowest credential barrier. If you understand business problems and how organisations actually work, integration consulting is where the money is going fastest.

The worst move is to wait until these roles are fully defined and crowded. The best career moves in AI right now are happening in the space between what is established and what is still being figured out.

That space is narrowing every month.

Reply and tell me which of these you're most interested in. I want to know which direction this community is thinking about moving.

— Prompt Guy

This newsletter goes out to over 95,000 people every week. If you have a product or AI tool you want introduced to this audience, reply with "sponsorship" and I will send you the details.