Skills-based job architecture
Skills sit on top of levels, not instead of them.
Short answer: A skills-based architecture helps a company hire, develop and move people by what they can do. It cannot replace job levels for pay, because pay comparisons and equal pay law look at the job, not at everything a person knows. Build skills on top of a level structure. Pay for scarce skills with premiums, not new levels.
Four hundred skills and no levels
Picture a 170-person company in Des Moines that builds underwriting software for regional insurers. Its new chief people officer has read a great deal about skills-based organizations. She proposes to retire the company's job levels and titles and move to a skills model. Every employee will have a skills profile, managers will rate each skill and pay will follow the skills a person holds and how well.
The first phase goes well. A vendor tool suggests about 400 skills from the company's job descriptions and résumés. Managers find the profiles useful for staffing projects. Two internal moves happen that would never have happened under the old job titles.
Then pay comes into it. An engineer with 40 skills on her profile asks why she earns less than a colleague with 25. A product manager asks whether learning a new analytics tool will lead to a raise. A sales engineer rated "advanced" in negotiation by one manager and "intermediate" by another asks which counts. And the finance lead asks how the company will compare its pay to the market when salary surveys are organized by job and level, not by skill.
None of these questions has an answer in the skills model. All of them had one in the level structure the company was about to retire.
This is a composite of a common situation at companies this size, not a client. The details and figures are illustrative.
Why skills-based architecture is hard to get right
Skills-based design sets two good goals against each other. The first is agility. Jobs change faster than titles. Describing work by the skills it needs makes it easier to hire people without the usual credentials, staff projects across teams and see where the company is short of capability. The second is fairness and stability in pay. Pay needs a unit that can be compared, priced against a market and explained. For most companies that unit is the job at a level.
Skills are spreading fast on the talent side. In a survey of US employers hiring new graduates, fielded in August and September 2025, 70% said they used skills-based hiring, up from 65% a year earlier. 81% used it when writing job descriptions [1]. A global skills survey found that 55% of organizations now map skills directly to jobs and 38% keep a single, company-wide skills library, both up from 2023 [2].
Pay has been slower to follow, for good reasons. Under US federal equal pay guidance, skill is measured by the experience, ability, education and training required to perform the job. A skill the job does not require should not be considered [3]. The EU pay transparency directive lists skills alongside effort, responsibility and working conditions as criteria for comparing jobs, which means skills enter pay through the job they belong to [4]. In both systems, the job is the anchor.
The market agrees. Salary surveys are organized by job and level. A compensation data provider reported in September 2026 that 48% of respondents found their market benchmarks did not reflect the skills roles now require. 23% of employers who had paid an AI skill premium now treat AI fluency as a baseline [5]. Skills move faster than the data used to price them.
For a company of 50 to 500 people, the question is rarely whether skills matter. It is where they should do the work. Skills are excellent at answering who could do this project, what this person should learn next and what the company will be short of in two years. They are poor at answering what this person should be paid compared with that one. A design that sends each question to the tool that answers it well avoids most of the trouble described below.
Two terms will help. A skills taxonomy is a list of skills a company uses to describe work. A level is a step in the size of a job, defined by scope, complexity and accountability, that applies across teams.
Five traps in skills-based architecture
"Replace jobs with skills"
What it looks like: The company retires job levels and titles and describes every person by a skills profile. Pay, careers and comparisons are meant to follow the profile.
Why smart teams fall into it: Levels feel rigid and titles feel political. A skills model looks modern and fairer, because it describes what people can do rather than what box they sit in.
What it costs: Without jobs, the company loses its comparison unit. It cannot say which people do equal work, cannot match roles to market data and cannot explain to an employee why a colleague earns more. Equal pay law still compares jobs by their content and requirements [3] [4], so the company ends up rebuilding jobs informally whenever a question arises. The move also tends to stall halfway. Managers keep using the old levels in their heads because they still need them to set offers and plan budgets, so the company ends up running two systems: The official skills model and the unofficial levels underneath. Most companies that use skills well keep levels and families as the frame and use skills to describe what each job needs. Our guide on job families and job levels covers how that frame is built.
"Pay for every skill someone has"
What it looks like: Pay rises with the number of skills on a person's profile or with each new certification, whether or not the job uses them.
Why smart teams fall into it: It rewards learning and sounds like the purest form of paying for skills. It also seems objective, because skills can be counted.
What it costs: Cost rises without a link to the value of the work. The people who gain most are those with the time and support to collect credentials. The legal position is direct. Federal guidance says possession of a skill not needed to meet the requirements of the job should not be considered when comparing jobs. A qualification can justify a pay difference only if it relates to job performance or benefits the business and the difference in qualification corresponds to the difference in pay [3]. Paying for skills a job does not use is hard to defend on both counts.
"The skills library will run itself"
What it looks like: A tool generates hundreds of skills from job descriptions and profiles. The library is launched and left to grow.
Why smart teams fall into it: Building a taxonomy by hand is slow. Software makes it fast and promises to keep it current.
What it costs: A library of hundreds of skills with no owner turns into a list of synonyms, overlaps and outdated tools. Managers rate the same skill under different names. Profiles stop meaning anything to the people who read them. Only 38% of organizations in one global survey keep a single, company-wide library [2], which suggests how hard this is. Tools that infer skills from documents add a further problem. They describe what people wrote, not what they do, so the profiles favor those who write long descriptions of their work. A small library tied to jobs, with someone responsible for it, is worth more than a large one nobody maintains.
"Skills data is objective"
What it looks like: Skills ratings, whether self-assessed or from managers, are treated as facts and fed straight into pay and promotion decisions.
Why smart teams fall into it: Skills sound measurable in a way performance does not. A list of proficiency levels looks like data.
What it costs: A skills rating is a judgment, with the same weaknesses as a performance rating. Different managers rate the same skill differently. Some people rate themselves generously and others modestly. Whoever is closest to the manager gets credited for the skill the team used. If those patterns line up with gender or background, a skills-based pay system carries them straight into pay, with an air of precision that makes them harder to question. Skills assessments need written criteria, evidence and the same check by group that any rating needs.
"The market will price the skills"
What it looks like: The company expects to set pay for each skill by looking up what the market pays for it.
Why smart teams fall into it: Job ads and data vendors increasingly quote premiums for specific skills, especially in AI.
What it costs: Advertised skill premiums are not what employers pay. A survey of more than 4,000 companies in February 2026 found employers paying around 10% above market for AI roles, mostly between 5% and 15%. 64% of compensation professionals were unsure what premium to offer [6]. Skill premiums also fade as skills spread [5]. Pricing every skill separately builds a pay structure on the least stable data available. A premium for a few scarce skills, on top of the job's range, with an owner and a review date, is easier to defend and to unwind. Our guide on paying for AI skills covers why the premium belongs on top of the band.
What a sound skills architecture looks like
A sound approach uses skills where they help most and leaves pay anchored to jobs. At 50 to 500 people, the approaches that hold up share these qualities.
Levels and families stay as the frame. Every job has a family and a level. Skills describe what the job at that level needs. The structure answers the pay questions and the skills answer the talent questions.
Skills are mapped to jobs first, people second. The library starts from what each job requires. People's profiles are then compared to the jobs they hold or want, which is where the value for development and mobility lies.
Pay follows the job's requirements. A person is paid for the job they do at the level it sits. Skills beyond that show up in development, in readiness for the next level and in internal moves, not in base pay.
A few scarce skills get premiums. Where a skill is hard to hire and changes what the job delivers, a premium on top of the range is paid, owned and reviewed. The list is short.
Assessment has rules. Proficiency levels have written definitions and examples. Ratings are checked across managers and by group, like any rating that affects pay. Where a skill can be shown through work, such as a project delivered or a system built, the evidence counts for more than the label.
The library is small and owned. A few hundred skills at most for a company this size, grouped by family, with one owner and a yearly review. Tools can suggest skills. People decide what goes in.
Skills describe the path to the next level. Each level lists the skills and scope that distinguish it from the one below. An employee can see what growth looks like in their family. A promotion is a decision about whether they now do the larger job, not about how many skills they have collected.
Jobs are re-leveled when their work changes. When a job's required skills shift enough to change its scope or complexity, it is re-leveled on that basis. Our guide on what to do when AI changes the work inside a job covers that judgment.
The calls only you can make
Where skills lead. Hiring, development, internal mobility or pay. Most companies get the most from the first three and keep pay anchored to jobs.
Which skills earn a premium. A short list chosen for scarcity and impact, or none. The list should be reviewed every year and allowed to shrink.
Who assesses skills. Managers, peers, the employee or a test. Each has costs in time and bias. Whatever you choose, apply it the same way across teams.
The size of the library. Broad and loose or small and precise. A small library is easier to keep honest.
What happens to titles. Titles can stay tied to levels or loosen into descriptions of skill. Loose titles feel modern and make the structure harder to read from outside.
What employees see. Whether people see the skills each level needs and their own profile against it. Visibility helps development and invites the question of what each skill is worth.
The data and the law
- Skills-based hiring. In a survey of 183 US employers hiring new graduates, fielded from August to September 2025, 70% reported using skills-based hiring, up from 65% a year earlier. 81% used it in writing job descriptions and 87% in interviewing [1].
- Skills mapping. A global skills survey for 2025 to 2026 found that 55% of organizations map skills directly to jobs, up from 47% in 2023. 38% keep a single, company-wide skills library, up from 30% [2].
- Benchmarks and baselines. A compensation data provider's September 2026 preview found that 48% of respondents said market benchmarks did not reflect the skills roles now require and 23% of employers who had paid an AI skill premium now treat AI fluency as a baseline [5]. These are the provider's own figures.
- What employers pay for AI roles. A survey of more than 4,000 companies in 133 countries, fielded in February 2026, found employers paying about 10% above market for AI roles, mostly between 5% and 15% [6].
- United States, equal pay. Under federal guidance, skill is measured by the experience, ability, education and training required to perform the job. Possession of a skill not needed for the job should not be considered. Education, experience and ability can justify a pay difference only if related to job performance or business needs and if the difference corresponds to the pay gap [3].
- European Union. Pay structures must allow work to be compared on objective, gender-neutral criteria including skills, effort, responsibility and working conditions. Relevant soft skills must not be undervalued (Article 4) [4].
This is general information, not legal advice. Rules and their national versions differ. Check the rules where you employ people.
Our view
We think skills are the most useful addition to job architecture in years and a poor replacement for it. They help a company hire beyond credentials, develop people deliberately and see what it lacks. They do not give the company a way to compare pay, price jobs or answer the equal pay question, because all three look at the job.
So we would keep levels and families, map skills to each job, use them for hiring and growth and pay premiums for a short list of scarce skills on top of the range. A company that does that gets the agility it wanted from a skills model and keeps the pay structure it still needs. It can also answer the question every employee eventually asks: Why am I paid what I am paid?
Where we help
Our job architecture work builds the levels and families first, then maps the skills each job needs and decides which few skills, if any, earn a premium. The starting point is job architecture, quoted within one business day. If a conversation would help first, Book a free call.
Sources
- NACE, Employer use of skills-based hiring practices grows (Kevin Gray, 12 January 2026; Job Outlook 2026, 183 employers, August to September 2025), checked 2026-10-30
- Mercer, Rebuilding reward and career frameworks based on skills (2025/2026 Skills Snapshot Survey), checked 2026-10-30
- US EEOC, Compliance Manual Section 10: Compensation Discrimination, 10-IV.E.2.a (skill) and 10-IV.F.2, checked 2026-10-30
- Directive (EU) 2023/970 on pay transparency, Article 4 (EUR-Lex), checked 2026-10-28
- Payscale, AI Workforce Impact Report preview, press release (24 September 2026; vendor survey), checked 2026-10-30
- Korn Ferry, Only modest pay bumps for AI skills (16 March 2026; survey of more than 4,000 companies in 133 countries, February 2026), checked 2026-10-28