We analyzed 254 job descriptions. Here's what employers actually demand.

Original data — July 2026. 254 real job postings from the public job boards of 39 companies, analyzed with the same competency engine Notch uses to score resumes.

Most advice about what to put on your resume is based on vibes: someone's opinion about what recruiters want, recycled across a hundred listicles. We wanted numbers instead. So we took Notch's competency engine — the framework that reads a job description and identifies what the role actually requires, and how essential each requirement is — and ran it over a corpus of real, public job postings.

No resumes, no surveys, no opinions. Just what employers wrote in their own job descriptions, extracted and counted.

The most-demanded skills

Ranked by the share of job descriptions that ask for them. "Importance" is how essential the posting makes each skill, on a 1–5 scale, where 5 means the posting frames it as a hard requirement.

The gap between #1 and #2 is the story. No hard skill comes close to cross-functional collaboration, because hard skills segment by role — SQL matters enormously in some jobs and not at all in others — while working across teams is demanded everywhere. It topped the list in every role family in the corpus: engineering, sales, finance, operations, marketing, all of them.

That has a direct resume implication most people miss. Candidates treat collaboration as filler ("team player") and cut it to make room for tools and technologies. The data says employers do the opposite: they name it, repeatedly, and flag it as essential — it was marked critical (importance 4 or 5) in 102 postings. The fix isn't to write the word "collaborative" on your resume; it's to make your bullet points name the teams you worked across and what shipped because of it. Our guide to writing evidence-driven bullet points covers how to do that without fluff.

AI is now a baseline skill — far beyond engineering

57% of postings named at least one AI-related competency: AI tools, prompt engineering, large language models, generative AI, machine learning. What's striking isn't the engineering number — you'd expect that — it's everywhere else:

Even if you exclude software engineering entirely, 54% of the remaining postings mention AI. Two caveats worth stating plainly: the small role families (Design at 8 postings, Product and People/HR at 12) carry wide error bars, and this corpus is drawn from technology companies, where AI adoption runs ahead of the broader market. But the direction is unambiguous — at these companies, "comfortable using AI tools in your work" is becoming what "proficient in Excel" was fifteen years ago: assumed everywhere, stated often.

If you use AI tools in your actual work — to draft, analyze, automate, or build — that belongs on your resume as concretely as any other tool. "Used ChatGPT" is weak; "built a prompt pipeline that cut report drafting from four hours to one" is a real bullet. As always: only if it's true and you can defend it in an interview.

"Communication skills" is not one thing

Job descriptions are more precise than resume advice gives them credit for. They rarely ask for "communication skills" in the abstract — they ask for a specific kind, and the kinds rank separately in our data: executive communication (13.4% of postings, importance 4.0), technical communication (8.7%), general communication (7.5%), stakeholder communication (4.3%), presentation skills. In total, 67% of postings name at least one.

Notice that executive communication — distilling work into what leadership needs to know — ties with Python for demand and carries a higher importance rating than SQL. It's one of the most valuable skills in the corpus and almost nobody puts evidence of it on a resume.

The tailoring move: identify which communication the posting asks for, and match your evidence to it. "Presented quarterly findings to the executive team" and "wrote the API documentation used by 200 integration partners" are both communication bullets — but they answer different job descriptions. This is exactly the kind of signal to extract when you analyze a job description before applying.

The hard-skill core: SQL, Python, Salesforce, Excel

Among tools and technical skills, four dominate: SQL (14.2% of postings), Python (13.4%), Salesforce (9.1%), and Excel (5.5%). SQL's position is worth dwelling on — it's the most-demanded hard skill in a corpus where engineering and data roles together make up only 23% of postings. Most of the postings that demand SQL aren't engineering jobs at all — operations and finance roles account for nearly half of SQL demand in the corpus, because "can you pull and interrogate your own data" has stopped being a specialist expectation.

And when Excel appears, it's serious: it carries an average importance of 4.4, and every one of its critical mentions is in the must-have register. Excel in 2026 isn't a filler line — the postings that name it (mostly finance and operations) treat it as core to the job.

When a skill appears, how negotiable is it?

Frequency tells you what's common; importance tells you what's fatal to be missing. These are the skills postings frame as most non-negotiable when they appear at all (minimum 5% of postings):

The pattern: role-defining skills are all-or-nothing. Financial modeling appears in only 5% of postings, but when it does, it's a 4.9 out of 5 — the job is financial modeling. Same for people management. Across the whole corpus, 65% of all listed competencies carry must-have framing rather than nice-to-have framing. Job descriptions are mostly requirements, not wish lists — which is why tailoring your resume to the specific posting beats sending the same resume everywhere.

What this means for your resume

Prove collaboration; don't declare it. The most-demanded skill in the corpus can't be claimed with an adjective. Name the functions you worked across, and attach an outcome: "Partnered with legal and engineering to ship SOC 2 compliance in one quarter."

Match the flavor of communication the posting asks for. Executive, technical, stakeholder-facing — the posting almost always tells you which one it means. Give evidence of that one, high on the page.

Surface your AI usage if it's real. A majority of postings at these companies now name AI-related skills. Concrete, defensible AI bullets are still rare on resumes — which makes them differentiating.

Treat named tools as pass/fail. When a posting names SQL, Salesforce, or Excel, the data says it usually means it (importance ~4+). If you have the skill, use the exact term — keyword matching is literal. If you don't, address the gap honestly or reconsider the application.

Assume everything listed is a requirement until proven otherwise. With 65% of competencies in must-have framing, the burden of proof is on treating something as optional — our guide to reading job descriptions shows how to tell the two apart.

Method

In July 2026 we collected 311 real job postings from the public Greenhouse job boards of 39 companies (including Stripe, Airbnb, Databricks, Cloudflare, GitLab, Figma, Reddit, MongoDB, Datadog, and Coinbase), sampling across departments on each board. We removed duplicate regional reposts of the same role, and capped every role family at 40 postings so no single function could dominate the aggregate — leaving N = 254.

Each posting was analyzed independently by Notch's competency extraction engine — the same framework Notch uses to score how well a resume matches a role, run here with no resume attached. For each posting it identifies the competencies the role requires and weights each from 1 (nice-to-have) to 5 (the role cannot be done without it), based on the posting's own language: must-have phrasing, repetition, "required" versus "preferred." Skill names are normalized so synonyms collapse into one bucket (e.g. "writing SQL" and "database querying" both count as SQL).

Corpus composition by role family: Sales 40, Software Engineering 38, Operations 31, Finance 27, Marketing 26, Data/Analytics 20, Customer Success 15, Product 12, People/HR 12, Design 8, Other 25.

Honest limitations: the corpus is drawn from technology companies with public Greenhouse boards, so it reflects hiring at tech companies, not the whole economy; small role families carry wide error bars; and competency extraction and weighting are model-assisted judgments applied uniformly across all postings, not human ratings. Every number in this article is computed directly from the corpus — nothing is estimated or adjusted.

Frequently asked questions

What is the most in-demand skill in job descriptions in 2026?

In this corpus, cross-functional collaboration — by a factor of five. It appeared in 73% of postings across every role family. The most-demanded hard skills were SQL (14%) and Python (13%).

How many job postings ask for AI skills?

57% of postings named at least one AI-related competency, including a majority of non-engineering postings. The corpus is tech-company hiring, where AI adoption runs ahead of the broader market — but within it, AI skills have clearly stopped being an engineering specialty.

Should I list soft skills like collaboration on my resume?

Not as a skills-list entry — self-declared soft skills carry no evidence. Demonstrate them in bullets instead: name the teams, the stakeholders, and what shipped. "Partnered with design and data science to launch X" proves collaboration; the word "collaborative" doesn't.

Can I see the underlying data?

The aggregate findings (skill frequencies, importance ratings, and corpus composition) are exactly what's published on this page, and the method is described above in full. If you're a writer or researcher and want more detail, email us at hi@notchresume.com.