What is a skills and job architecture?
A skills and job architecture is a structured framework that defines the skills required for each role and level, then organizes those roles into job families and career paths. It gives every talent decision, from hiring to mobility, learning and workforce planning, the same foundation.
It has two layers. The skills architecture describes what each role requires. The job architecture groups roles that share common skills and maps the bridges between them.
Most organizations already hold this information. It's just scattered across thousands of job descriptions and postings, written in inconsistent terms and often in several languages. Nobody can read it all, so people decisions end up resting on job titles alone.
Skills taxonomy, ontology or framework: what's the difference?
A skills taxonomy classifies skills into categories, a skills ontology maps how skills relate to each other and to roles, and a competency framework defines the proficiency levels expected. They work together: a solid architecture starts with a taxonomy at minimum, then adds the other two.
| What it is | What it's for | Example | |
|---|---|---|---|
| Taxonomy | Skills sorted into categories | A shared list to identify, assess and track | Data science sits under technical skills and includes Python, data analysis and machine learning |
| Ontology | Links between skills, roles and tasks | Matching and mobility | Data science connects to statistics and Python, and to roles like data scientist or AI engineer |
| Framework | Proficiency levels and their criteria | Assessment and development | A leadership framework breaks strategic thinking into observable behaviors at each level |
Why does a skills and job architecture matter now?
Because skills are changing faster than job titles. According to the World Economic Forum's Future of Jobs Report 2025, workers can expect 39% of their existing skill sets to be transformed or become outdated by 2030, and 63% of employers name skills gaps as the biggest barrier to business transformation.
Without a shared architecture, every skills project starts from scratch and builds its own list. With one, four uses run on the same foundation:
- Future-proof workforce skills: spot gaps early and plan upskilling or reskilling before they become urgent.
- Boost engagement and retention: when career paths are defined in skills, people know what to learn to move up.
- Open hiring to skills: assess what candidates can do, including those from non-traditional backgrounds.
- Align talent with strategy: a new market, product line or regulation demands specific skills; know which you have and which to build or hire.
How do you build a skills and job architecture in 9 steps?
Start from strategy, map what you have, define core skills, then group jobs by skills and build careers, hiring and culture on top. These 9 steps work with or without a vendor, whether you're starting from scratch or refining what exists.
- Assess organizational needs and goals: where is the business going, and which skills will matter in the next 3 to 5 years?
- Assess the skills you have: self-assessments, manager reviews, skills tests and performance reviews, centralized in one tool.
- Define core skills: foundational skills for everyone, role-specific skills for each job, organized into a taxonomy.
- Group jobs by skills, not titles: job families built on shared or complementary skills.
- Build career paths and learning plans: key skills per role, expected proficiency, and the development needed to get there.
- Use technology where it helps: reading documents, suggesting skills, tracking progress.
- Embed skills in hiring: demonstrated skills over years of experience in a specific title.
- Build a skills-based culture: continuous learning, mentoring, recognition and cross-functional collaboration.
- Measure and adjust continuously: an owner for each part of the architecture, and regular reviews.
Tip: much of this information already exists. Job descriptions and postings already describe the tasks of each role.
How do you build a skills architecture from your existing documents?
At 365Talents, we built it in 8 phases: choose your sources, clean up the role catalog, map each document to a role by its content, extract skills, consolidate the vocabulary, compose each role's list, enrich it if needed, then have your experts review it. Every skill stays traceable to its source.
- Sources: your job descriptions, filled positions and job postings, plus market data if you choose. You decide the mix.
- Roles: consolidate titles (catalogs often carry far more titles than real roles), fill missing descriptions, place every employee under exactly one role.
- Mapping: each document is matched to a role by what it describes, not by its title.
- Extraction: only the tasks are kept and rewritten as skills. What isn't in the text doesn't come out of it.
- Consolidation: variants ("Preventive maintenance", "PM Scheduling", "Mantenimiento Preventivo") merge under one name, and you choose how granular skills should be.
- Composition: about a dozen skills per role, balancing today's organization with tomorrow's needs.
- Enrichment (optional): definitions, categories, labels in each language, proficiency levels.
- Review: your experts keep, rename, reject or add each proposed skill.
Example. For a line maintenance technician, predictive maintenance didn't appear in any internal document. But it was growing fast in similar roles across the industry, and two production lines were already piloting machine condition monitoring. The maintenance manager added it: an architecture built only on existing documents would have missed it.
What role should AI play in a skills architecture project?
AI handles the reading, which is most of the work; your teams keep the decisions. It can process thousands of documents in several languages, map each one to the right role, extract skills, group different wordings of the same skill and suggest emerging skills from market data.
| AI handles | Your teams decide |
|---|---|
| Reading thousands of job descriptions and postings | Which sources are relevant |
| Mapping each document to a role | How much weight to give external data |
| Extracting skills from the text | How to group roles |
| Grouping different wordings of the same skill | Which skills each role requires |
| Drafting missing descriptions | Which suggested skills actually fit a role |
| Suggesting emerging skills | Which emerging skills matter for the business |
Before choosing a tool, ask three questions: can you trace every skill back to its source? Who reviews the draft, and how long does it take? Can the framework be exported to your other HR systems?
How long does it take to build a skills and job architecture?
For a typical scope of about 350 roles, expect around 8 weeks with 365Talents to a first reviewed draft and 14 weeks to go-live. Larger organizations usually extend the review phase, not the build.
On the HR side, the effort comes down to three things:
- An export, not a project: your job catalog, filled positions and job descriptions, typically one day of work for an HRIS specialist.
- A few key decisions, made in a few workshops: which sources to use, how to group roles, how granular skills should be.
- Reviewers who know the work: about a dozen skills to review per role, roughly half a day for 20 roles.
No. But without it, the architecture only contains what you've already written down, and usually misses recent and emerging skills.
Descriptions can be generated from job titles, organizational context and similar roles, then corrected by your teams. The starting point is less precise, so review takes longer.
It's normal. Consolidating titles is part of the method, and many organizations have never deduplicated their catalog.
Yes. Each skill can be published with a validated label in every working language, while still counting as one skill in your analytics.
No. Levels are essential for assessment campaigns, but matching, mobility and coverage analysis work without them.
Go further
The full playbook covers the 9 steps, the 8-phase method and one role followed end to end: Download the playbook.
