In-demand skills in 2026: what live job postings ask for, by function

Every list of in-demand skills you have read was written from a survey or from the author’s head. This one is read off live job postings: for each job function, the share of active postings that name a skill, extracted from the title and description against a curated lexicon. The shares are cross-sectional, refreshed with the corpus, and reproducible through the API.

What are the most in-demand skills right now?

Across every function the most-named skill is communication, followed by compliance and leadership; the first hard skill depends on the function. In Engineering it is Python, named in 19.2% of postings; in Finance it is Excel; in Healthcare it is patient care; in IT and Security it is cybersecurity. The tables below give the top skills per function.

A note on reading them. The lexicon includes soft skills, so “communication” leads everywhere and tells you little. The signal is in the hard skills and in the gaps between functions: a skill near the top of one function and absent from another’s top 25 is a hiring-market fact, not an opinion.

What “in demand” means in postings data

In this dataset a skill is in demand when a large share of postings in a function name it. That is an employer ask, not a hire, and it mixes two different things: skills that are common because every posting lists them, and skills that are scarce and therefore spelled out. Communication is the first kind. Kubernetes is the second.

The share also depends on how the skill is written. A lexicon matches the phrases employers use, so a skill with one canonical name (Excel, CPR, Salesforce) counts cleanly, while a skill with many names is split across slugs and each slug looks smaller than the skill. Survey-based lists do not have this problem and have a worse one: they measure what respondents say they value, not what they wrote into a requisition. Read the tables as what employers ask for in writing, function by function, and nothing more.

In-demand skills by job function

What stands out per function

• Engineering. Python is the first hard skill at 19.2%, ahead of CI/CD at 13.1% and AWS at 11.5%. Lean manufacturing at 13.6% is a reminder that the function includes mechanical and industrial engineering, not only software.
• Sales. CRM is named in 25.0% of postings and Salesforce by name in 9.4%; lead generation, negotiation and account management sit close together at 15.9%, 15.8% and 15.8%.
• Marketing. SEO at 13.7% and Google Analytics at 10.2% lead the tool skills, with Canva at 10.1% ahead of Google Ads at 8.2%.
• Finance and Accounting. Excel is named in 38.9% of postings, the CPA in 15.0%, and SAP in 9.3%.
• IT and Security. Cybersecurity at 21.1% is the first hard skill; Azure at 16.3% edges AWS at 14.4%, the reverse of Engineering, and identity and access management reaches 12.8%.
• Healthcare. Nursing at 43.3%, patient care at 36.9% and CPR at 31.0% dominate; EMR systems appear in 8.6% and HIPAA in 4.8%.

Two functions are published in the dataset but not charted here: Human Resources, where recruiting, payroll and HRIS lead the hard skills, and Operations, where project management and lean manufacturing do. Both are in the JSON behind this page and in the skills taxonomy post.

Are AI skills in demand?

Yes in Engineering, where an explicit artificial-intelligence skill tag reaches the top 25; not measurably elsewhere. In every other function no AI tool or framework slug reaches the twenty-fifth skill, which is the cut-off of this dataset, so the honest answer for those functions is “below 3.9% of postings”, not zero.

The pay side of the same question is on the Labour Market Pulse, which publishes the share of postings mentioning AI tooling per occupation and the gap in median advertised pay between AI and non-AI roles. Specific vendors are better queried than ranked: skills_or accepts slugs such as openai, langchain and pytorch, and technology_slugs on every returned row lists the tools the posting names.

How this changes by board

Skills are extracted from the description, so the first suspect when a ranking shifts between boards is description length. The data says it is a small effect: the share of scored postings the ghost scorer flags for a thin description, under 400 characters, is low on every collected board, 1.3% across all ten, and highest on Lever.

What moves the ranking is the mix. Each board carries its own blend of functions, industries, countries and seniority levels, so the same skill has a different share on each. A ranking built from one aggregator inherits that aggregator’s mix; a ranking built from ATS feeds such as Greenhouse, Lever and Ashby leans toward the companies that use those platforms. The tables above merge all ten collected boards. When you build a skills ranking for one function, filter by source_or and compare the boards separately before merging them, and read the per-board field coverage in the job scraping explainer first.

How to pull the same cut through the API

There is no function filter on search, so scope by occupation code and then read keyword_slugs and technology_slugs off each row. This pulls recent ICT professional postings naming Python or Kubernetes:

curl -X POST https://api.jobspipe.dev/v1/jobs/search \
  -H "Authorization: Bearer $JOBSPIPE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "occupation_code_or": ["25"],
    "skills_or": ["python", "kubernetes"],
    "posted_at_max_age_days": 30,
    "limit": 100
  }'

Count slugs across the pages and divide by the rows returned, and you have the share table above for any slice you like: one country with job_country_code_or, one seniority with job_seniority_or, or one company with company_name_or. The ISCO-08 post lists the occupation prefixes.