SFIA 9 framework alignment

What this page is

A public-facing record of how Capable represents the SFIA 9 framework, what we cover, what we deliberately don't, and the boundaries we hold against AI improvisation. This page is the buyer's honest read on framework fidelity, and it is the SFIA Foundation's reference if it ever needs to audit how we present the framework inside the platform.

The platform is built around the SFIA 9 release published by the SFIA Foundation in October 2024. Everything below cites that release directly.

Version and source

• Framework: SFIA 9 — published 2024-10, current at time of writing.
• Source file: sfia-9_current-standard_en_250129.xlsx, authoritative SFIA Foundation release, retrieved 2025-01-29.
• Locked into the platform: The framework is bundled with the application as a content-addressable JSON resource. Every assessment claim cites the framework version it was generated against, so a claim made today is reproducible against the same source a year from now.
• Update path: When the SFIA Foundation cuts a new release (v10 / point releases / errata), the framework JSON is refreshed through the same versioned-content pipeline that governs every authorable shape on the platform — draft, review, publish, with a cryptographic payload hash recorded on every version.

Coverage summary

Counts below are taken from the bundled framework data (/sfia9_framework.json, statistics block):

Surface	SFIA 9	Capable
Categories	6	6/6
Subcategories	22	22/22
Skills	147	147/147
Levels of responsibility	7	7/7
Generic attributes	16	16/16 (4 Attributes + 12 Business skills / Behavioural factors)
Skill-level descriptions	672	672/672

The 26 new skills introduced in SFIA 9 — including AIDE (AI and data ethics), NFTS (Non-functional testing), RMGT (Records management), ANCC (Analytical classification and coding), and the broadened PEDP (Information and data compliance) — are all present and assessable.

Generic attributes

SFIA 9 restructured the generic attributes into two categories of unequal status:

Attributes (4). The load-bearing dimensions of the responsibility ladder. Every skill descriptor is anchored to them:

• AUTO — Autonomy
• INFL — Influence
• COMP — Complexity
• KNGE — Knowledge

Business skills / Behavioural factors (12). The supporting dimensions that colour how responsibility is exercised at each level:

• COLL — Collaboration
• COMM — Communication
• IMPM — Improvement mindset
• CRTY — Creativity
• DECM — Decision-making
• DIGI — Digital mindset
• LEAD — Leadership
• LADV — Learning and development
• PLAN — Planning
• PROB — Problem-solving
• ADAP — Adaptability
• SCPE — Security, privacy and ethics

Every piece of evidence captured during an assessment is tagged with the SFIA 9 attribute it bears on. The split between the two categories is preserved in reporting — Attributes drive the level claim; Business skills / Behavioural factors describe how that level shows up in practice.

How claims trace to descriptors

Three load-bearing properties hold for every assessment claim the platform produces:

1. Verbatim descriptors, not paraphrases. When the platform shows a SFIA level description, it renders the exact SFIA 9 text from the bundled framework JSON. The AI layer does not "summarise" or "improve" descriptors. If a phrasing is shown alongside the official descriptor, it is clearly labelled as a paraphrase for participant readability and the verbatim descriptor is one click away.
2. Immutable version binding. Every response a candidate gives carries the content_version_id of the exact item payload they saw. The item payload is sha256-hashed and write-protected by a Postgres trigger after publish. A misbehaving migration cannot silently mutate a published item; the database refuses.
3. Traceable to the SFIA short-code URL. Every attribute and level surface in the platform links to the corresponding sfia-online.org page (e.g. https://sfia-online.org/en/shortcode/9/AUTO for Autonomy). The platform is a presentation layer; the SFIA Foundation is the canonical source.

The full per-version audit trail is exposed through the Govern audit log surface, filterable by content_version_state_changed events.

What we do not do

In the spirit of the SFIA Foundation's published position on high-risk AI systems in HCM, the platform commits to a small set of "we don't"s:

• We do not let the AI improvise framework content. Descriptors are rendered from the framework JSON, not generated.
• We do not re-norm SFIA levels. Our scoring band edges follow SFIA's published level definitions. If we surface a derived threshold (e.g. "competent for this role"), it is anchored to a named level and explained.
• We do not copy SFIA descriptor content onto our marketing surfaces. The framework lives behind the assessment surface, under licence, with attribution.
• We do not claim endorsement that we have not been granted. If Capable is listed as a SFIA Foundation Accredited Partner, this page will say so; until then, this page says that we are a tool built on the SFIA 9 framework, which is a fact about the framework being open for use under licence.

Attribution and licensing

SFIA® is a registered trademark of the SFIA Foundation, protected in more than 35 countries. SFIA framework content is © SFIA Foundation, used under licence.

Capable's product name does not incorporate "SFIA". The Capable brand is more prominent than the SFIA brand across every surface, per the SFIA brand and trademark rules. Customers using Capable in their own organisations are responsible for obtaining their own SFIA licences where required; see the SFIA Foundation's licence chooser.

If you are with the SFIA Foundation and would like to discuss commercial-partner status, please reach out via contact — we are happy to walk through the framework fidelity surface live.