SFIA 10 consultation: AI and professional work
Background
AI is changing professional work. For SFIA, this raises two related areas to consider:
- the professional skills needed to develop, deploy, manage and govern AI systems
- how AI changes the way existing professional skills are applied across other roles and disciplines.
For SFIA 10, we are considering how the framework should reflect both areas. This includes whether new or changed skills are needed, how SFIA should describe the everyday capability to use and supervise AI, and how it should reflect responsibility and accountability in AI-enabled work. AI was the theme raised most strongly during the SFIA 10 consultation workshops.
Employers are already using SFIA 9 to develop their professional workforce while AI is changing how work is carried out. The resources presented here can be used with SFIA 9 now. Experience and feedback from their practical use will also inform the development of SFIA 10.
The consultation on SFIA 10 remains open. We welcome feedback from across the SFIA and wider professional skills community.
SFIA is a trusted, internationally adopted framework for defining the skills and competencies needed to deliver, manage and protect digital capabilities, including in areas such as AI, data science and cybersecurity. SFIA 10 will build on the established framework and the experience of its global community. It will integrate the implications of AI into SFIA as a whole, rather than treating AI as a separate domain requiring a new framework.
Overview
Understanding AI and SFIA 9
Helping users now and informing SFIA 10
AI impact index
Explore how AI may affect each SFIA skill, what remains a matter for professional judgement and how the impact changes across responsibility levels.
About the AI impact indexAI concepts register
Start with familiar AI terminology and find the SFIA skills that people may draw on when working with each concept.
About the AI concepts registerExploring changes for SFIA 10
Prototype content and areas under investigation
Working with AI
Explore a cross-cutting description of the capability to use AI responsibly, evaluate its outputs and provide appropriate supervision and oversight.
About the Working with AI prototypePossible new or changed SFIA skills
See the areas of AI-related professional work being examined to determine whether existing skills, additional guidance or new SFIA skills are needed.
See the areas under investigationExplore the impact of AI with these complementary resources
We are making three complementary resources available now for use, exploration and feedback.
We are making three complementary resources available now for use, exploration and feedback:
- AI impact index: Explore how AI may affect each of the 147 skills in SFIA 9.
- AI concepts register: Start with familiar AI terminology and discover the SFIA skills associated with it.
- Prototype Working with AI attribute: Explore a cross-cutting, seven-level description of the capability to work with AI effectively and responsibly.
This work was initiated through the research and consultation informing SFIA 10. It is helping us examine how AI is changing professional practice, whether SFIA continues to describe that practice effectively and where changes, clarification or additional guidance may be needed.
Why we are publishing this now
We could have kept this work internal until SFIA 10 was published. We have chosen to release it now for two reasons:
- It is useful now. The material has practical value for current SFIA users, independent of the decisions eventually made for SFIA 10.
- Practical use will improve the work. Feedback from applying these resources to real roles, teams and organisations can provide stronger evidence for the SFIA 10 review.
This reflects our broader approach to developing SFIA: sharing emerging work, reasoning and evidence, not only the eventual outcome.
1. AI impact index
How AI affects each of the 147 skills in SFIA version 9
How AI's impact on SFIA is being assessed
We've systematically reviewed all 147 SFIA version 9 skills to understand how AI is changing professional practice in each one. Rather than starting from the assumption that AI needs its own skill or section, we've asked the same question of every skill: what might AI change, and what remains a matter for human professional judgement and accountability?
This work gives the SFIA Design Authority a systematic and traceable starting point for deciding what, if anything, should change in SFIA 10.
Let us know: Is this a useful resource? What else would be useful?
AI impact index - How AI affects each of the 147 SFIA-9 skills
How to use the AI commentary for SFIA skills
2. AI Concepts register
Search everyday AI terminology and see which SFIA skills it draws on
64 AI concepts indexed, 95 distinct SFIA skills referenced, 314 concept–skill relationships across 7 categories:
- Data foundations for AI
- Core AI & machine learning techniques
- Building & integrating AI systems
- Assuring AI quality, performance & security
- Governing, protecting & regulating the use of AI
- Working with AI
- AI Strategy, value & delivery
PROTOTYPE AI concepts register
3. Cross-cutting Working with AI
Working with AI (WWAI) describes the everyday capability to use AI tools effectively and responsibly,
critically evaluate their outputs and provide appropriate supervision and oversight.
It applies across all types of roles. Like SFIA’s generic attributes and behavioural factors, it is cross-cutting rather than associated with a particular professional discipline. It uses the SFIA levels of responsibility to describe how working with AI changes as responsibility, influence and accountability increase.
Working with AI complements SFIA’s existing generic attributes and behavioural factors. It focuses on effective and responsible practice rather than particular techniques and does not prescribe specific tools, technologies or methods.
Its development has considered activities such as:
- using prompting techniques
- critically evaluating and validating AI-generated outputs
- supervising and overseeing AI systems
- identifying and managing problems such as agent drift
- undertaking roles in which these activities form a significant part of the work.
Prototype status: Working with AI has been released as a prototype. It is not dependent on SFIA 10, so organisations and individuals may find it useful now.
- Its eventual positioning within SFIA will be considered as part of SFIA 10.
- In the meantime, we encourage users to apply the prototype and provide feedback as we test and refine it.
Let us know: Is this a useful resource? What else would be useful?
How responsibility progresses
Working with AI (WWAI) means using AI responsibly and critically within agreed boundaries for work, risk, decision-making and accountability.
Across the SFIA levels, this progresses from using approved AI for simple assigned tasks, through taking responsibility for one’s own AI-enabled work and enabling others, to establishing effective practices within an area of responsibility.
At the higher levels, it includes leading organisational implementation and setting the overall approach across strategy, operating models, risk, decision authority and accountability.
What about 'prompt engineering'
Terms such as prompting and prompt engineering describe some of the techniques used to provide AI with instructions, context, constraints or source material. WWAI describes the underlying professional behaviour rather than prescribing particular interaction techniques.
WWAI deliberately describes behaviours that are durable and should survive changes in:
- AI products
- interaction methods
- technical architectures
- terminology
- specific techniques.
It avoids confusing technique with behaviour. Someone could be highly skilled at constructing or re-using others' prompts but poor at judging whether the resulting output should be trusted or used.
It avoids implying that prompting is universally required.
SFIA generally describes capability and responsibility rather than prescribing specific techniques.
Everyday use and specialist work
It is helpful to distinguish between two situations.
Using prompts as part of ordinary professional work
A person may provide instructions, context or source material to help an AI tool perform a task. This is covered by Working with AI alongside expectations for checking outputs, protecting information and exercising professional judgement.
Designing prompting and context arrangements as specialist work
Some practitioners design, test and maintain reusable prompts, system instructions, context-management arrangements or AI-enabled workflows. This can be substantial professional work.
The relevant SFIA skills depend on what the practitioner is doing and the responsibility they hold. The work may draw on skills concerned with systems design, software development, machine learning, data science, user experience, testing or other areas.
How we consider emerging AI skill labels
New terms such as prompt engineering, harness engineering and loop engineering may identify important changes in professional work. We take these developments seriously, but the appearance of a new job title or the addition of the word “engineering” does not automatically establish a new discipline or SFIA skill.
For SFIA 10; we examine the work behind the term:
- What activities and outcomes does it describe?
- What responsibility and accountability do practitioners hold?
- Is it already covered by one or more SFIA skills?
- Is it distinct enough to be assessed separately?
- Can it support meaningful progression across responsibility levels?
- Is it likely to remain useful as technologies and terminology change?
The result may be a new search term, additional guidance, an update to an existing skill, or a group of skills presented through a View or role profile. A new SFIA skill is appropriate only where the work has sufficient scope, distinctiveness and durability.
Adding skills too quickly risks duplicating existing content, fragmenting professional practice and tying SFIA to short-lived terminology.
4. Exploring possible new or changed SFIA skills
We are examining whether some areas of AI work require new SFIA skills or changes to existing skills.
The Working with AI prototype describes a cross-cutting capability relevant to many different roles. In parallel, we are examining whether some areas of AI work require new SFIA skills or changes to existing skills.
The appearance of a new term, tool or technique does not automatically mean that SFIA needs a new skill. For each area, we will consider whether:
- the work is already covered but needs to be easier to find
- an existing skill or its guidance needs clarification or updating
- the area is better addressed through a SFIA View or role profile
- there is a distinct and durable area of professional work that requires a new skill.
Current areas under investigation include:
- managing the AI lifecycle, including MLOps and model observability
- AI evaluation and ongoing assurance
- AI safety and assurance by design
- agent identity, delegated authority and accountability for actions
- explainability and interpretability
- retrieval, enterprise search and retrieval-augmented generation
- designing human interaction with AI systems.
These are areas for investigation, not proposed skill names or commitments to add new skills. Some overlap with one another and with existing SFIA skills.
Before proposing a new skill, we will look for a recognisable area of professional work that is not adequately described elsewhere. It should be distinct from adjacent skills, support a meaningful progression across responsibility levels and remain useful as technologies and terminology change.
SFIA 10 Consultation - AI themes and topics
Addresses the everyday capability to use AI tools effectively, evaluate their outputs, and supervise their operation.
Working with AI as a theme for SFIA 10
AI is changing how a great deal of professional work is done, not only the work of building AI systems. For SFIA 10 we are looking at how the framework describes what people are accountable for as AI becomes part of everyday practice, how it might recognise the everyday capability to use and supervise AI tools, and how our guidance reflects the way work is actually carried out. This is the area the consultation raised most strongly.