Skip to content

Skills

Which tech skills are paying off

Some skills are getting more valuable fast and some are being automated away. Our read on both, the new roles we're seeing, and a step-by-step plan for your job. Last reviewed September 2026.

Rising demand

  • AI/LLM engineering (RAG, agents, evals)

    LinkedIn says AI engineer is the fastest-growing job title in the US in 2026. Indeed found AI in 6.3% of all postings in August 2026, and Robert Half puts starting ranges at $134K to $193K.

  • AI workflow redesign and adoption

    Employers say 'AI adoption and automation workflows' is the hardest capability to find in 2026. If you've owned one real workflow redesign, that counts for more than any certificate.

  • AI literacy for non-engineers

    Prompting, AI-assisted workflows and tool orchestration now show up in job descriptions in every function. PwC measured a 56% wage premium for roles that require AI skills.

  • Cybersecurity & AppSec

    AI widens the attack surface faster than defenses keep up, and security roles have held their demand through every layoff wave since 2022.

  • Cloud & platform engineering

    AI workloads run on cloud infrastructure, so FinOps and platform skills ride the same capex wave as AI. Azure proficiency is on employers' hardest-to-fill list for 2026.

  • Data engineering & analytics engineering

    Every AI project starts with data plumbing. Data engineering is a top-three in-demand role for 2026, and Kafka and Databricks are named among the scarcest skills.

  • AI product management

    PMs who can scope, evaluate and ship LLM features are hard to find. Demand for ordinary PMs is flat or falling.

  • Domain + tech hybrids

    Pair domain knowledge in health, fintech, legal, energy or industrial work with technical skill and you have the profile that's hardest to automate or offshore.

  • AI safety, evals & governance

    The EU AI Act and enterprise adoption are creating compliance work that barely existed in 2023.

Declining demand

  • Manual QA testing

    More of this role has been automated away than any other this cycle. Postings are down sharply as AI-assisted and automated testing spread.

  • Routine content writing

    Volume copywriting was the first market where generative AI visibly pushed prices down, back in 2023.

  • Basic data entry & reporting

    This is the IBM category: back-office roles that shrink through attrition, with no layoff announcement.

  • Tier-1 customer support

    AI agents now handle a large share of ticket volume at major SaaS companies (Salesforce, and Klarna with some caveats).

  • Traditional recruiting coordination

    Scheduling, sourcing and screening are now heavily AI-assisted, and recruiting was the hardest-hit function of 2023 to 2024.

  • Pixel-only production design

    Asset production is shrinking, and design demand is moving up to systems, research and product thinking.

When we say declining, we mean the basic version of the skill pays less than it used to. There's usually a related version that pays more, and the plans below show how to get there.

New kinds of jobs we keep seeing

Who's hiring for them →

AI / ML engineer

Pretty much every company that cut staff in 2026 is also building AI features, and the people who can ship them are the hardest hires going.

LLM integration · RAG and agents · Evals

Data engineer / analytics engineer

Most AI projects stall on messy data. Someone has to build the pipelines and warehouses first, and that's where the budget goes.

SQL and dbt · Kafka · Databricks / Spark

Security engineer

Every new AI tool is one more thing to attack. Security hiring held up through every layoff wave since 2022.

AppSec · Cloud security · Identity

Platform / DevOps / cloud engineer

AI runs on infrastructure somebody has to build and keep running, and keeping the cloud bill down (FinOps) has become a job of its own.

Kubernetes · Terraform · FinOps

Forward-deployed / AI solutions engineer

The labs and their customers need people who can get a model working inside an actual business. Think half engineer, half consultant.

Prompting and evals · Systems integration · Customer discovery

AI product manager / AI enablement lead

Someone has to pick which work AI should take over, check whether it actually did, and retrain the team. Demand for general PMs is flat. This version is growing.

Workflow redesign · Eval design · Change management

AI governance, evals and safety

The EU AI Act and nervous risk teams are creating compliance and testing jobs that didn't exist in 2023.

Evaluation methodology · Risk frameworks · Policy

Domain expert who builds with AI

If you know finance, health, law, energy or manufacturing and you're good with AI, you're very hard to automate or offshore.

Regulated-domain depth · AI workflow design · Data literacy

A plan for your role

Pick your job below. We'll tell you plainly what's changing and what we'd do about it.

Let AI write the routine code while you own the system

Software Engineer

The shift: AI now writes a large share of routine code. We don't think that means the end of engineers. What it does mean is that the market needs fewer people whose main value is turning out straightforward code, and more people who can own a system, review AI output with a critical eye and turn a vague problem into an architecture.

  1. Get fast with AI coding tools

    Use Copilot, Claude or Cursor every day until you're at 2 to 3x your old throughput. More and more teams measure output against an AI-augmented baseline, so being fast without AI doesn't clear the bar anymore.

  2. Ship one LLM feature to production

    RAG over internal docs, an agent for a support workflow, structured extraction, anything that actually runs in production. That single line on a resume changes response rates.

  3. Go deep where models are weak

    Distributed systems, performance, debugging production incidents, security. AI is weakest in exactly the places where the stakes are highest.

  4. Take ownership beyond your tickets

    Lead an epic, run incident retros, work directly with product and customers. Coordination and judgment are the parts of the job that last.

Tools to know

GitHub CopilotClaude CodeCursorLangChain / LlamaIndexOpenAI / Anthropic APIs

Not sure which of these applies to you?

The calculator takes two minutes and points at the moves that matter most for you.

Try the calculator