Agenda item 3: AI and the workplace

SPC Members are asked to agree their preferred option. 

• To await the report of the WPC in November and determine next steps at that point; or
• To agree to establish a sub-group of the SPC and build on the work of WPC for the benefit of all sectors involved in the SPC.

1. The first meeting of the Social Partnership Council (SPC) on 1 February 2024 highlighted the impact of Artificial Intelligence on the workforce as a potential topic for the further exploration by the Council. A proposal was made that an SPC sub-group could be formed to consider this issue further.
    
2. The SPC is asked to consider how to develop this work. 

3. The item was raised at the SPC in February by Ruth Brady (RB), Regional Secretary GMB. Ruth is also a member of the Workforce Partnership Council (WPC) and Chairs a WPC working group that is considering the impact of AI on the public sector workforce.
    
4. The WPC is a social partnership structure comprised of trade unions, employers, and Welsh Government. It covers the devolved public services in Wales and is a forum for cross-public services workforce matters. 

5. The working group is made up of trade union representatives, employer representatives, Welsh Government policy leads, and additional co-opted members with technical expertise.
    
6. The working group has drafted and agreed an assessment of the public sector workforce implications, considering both the threats and opportunities that AI may present.
    
7. To help strengthen the communication across national level social partnership structures and embed understanding of the various structures, a copy of the assessment document and explanatory note will be sent to the following forums for their awareness:

• NHS Wales Partnership Forum
• Local Government Joint Council of Wales
• Devolved Public Sector Group
• Schools Social Partnership Forum 

8. A copy of the assessment document is attached for reference (annex A).
    
9. The agreed objectives of the working group are included at annex B.
    
10. The working group has emphasised the urgency in developing a response to this fast-moving agenda. To aid this development, the group has identified two broad work themes to achieve its’ objectives:

• Best practice, guidance, and progress on existing WPC digital principles;
• Benchmarking the understanding of AI and considering the benefits of AI.

11. The business of the WPC is progressed between Council meetings by the Joint Executive Committee (JEC). The next meeting of the JEC will take place 18 September where a first draft report will be considered. The WPC will then take a decision on the final report 13 November. 

12. Section 8 of the Social Partnership and Public Procurements (Wales) Act allows for the SPC to establish relevant subgroups. A subgroup of the SPC may (a) carry out any function under section 1 of the Act delegated to it by the SPC; and (b) help the SPC carry out its functions in any ways specified by the SPC.
     
13. There will be a strong synergy between the themes being looked at by the WPC AI working group and an SPC sub-group, if established.  However, whilst the WPC working group has a wide membership, it largely consists of public sector employers.  
     
14. An SPC group would be able to look at the issue through a wider lens and develop an understanding of the needs of the further and higher, private and third sectors, the levers available to embed the responsible use of AI, and any specific barriers these sectors may experience. The group could consider materials available across all identified sectors, in addition to the materials and outputs from the WPC AI working group.
     
15. Membership of the subgroup would not necessarily come from members of the SPC but would be proposed by them. A subgroup could also bring in expertise as required to further supplement any findings or advice it provides to the SPC and could include members of the WPC working group.

The SPC Members are asked to agree their preferred option. 

• To await the report of the WPC in November and determine next steps at that point; or

• To agree to establish a sub-group of the SPC and build on the work of WPC for the benefit of all sectors involved in the SPC.

16. If an agreement is reached to establish an SPC sub-group, the Secretariat will invite volunteers and suggestions for membership. 

17. If members would benefit from further information on AI there are a number of useful documents.  

Guidelines for AI and Shared Prosperity - Partnership on AI

Home | Centre for Digital Public Services (gov.wales)

A snapshot of workers in Wales’ understanding and experience of AI | TUC

AI Principles Overview - OECD.AI

AI Ethics and Governance in Practice | The Alan Turing Institute

A guide to using artificial intelligence in the public sector - GOV.UK (www.gov.uk)

Algorithmic Transparency Recording Standard - GOV.UK (www.gov.uk)

Going public | Ada Lovelace Institute

1. In some sectors, AI could automate or augment many tasks and roles that are currently performed by public sector workers, such as data entry, processing, analysis, customer service, and compliance. This could lead to job displacement or loss for some workers.
    
2. Overall, AI will lead to a shift in the skills required by the public sector workforce, but the long-term implications will need to be continuously under review. AI is likely to fuel demand for currently scarce skills such as data science and/or coding/developers, all of which have a high market premium, as well as new skills to use AI tools such as prompting. The technologies also present new governance challenges, such as transparency and explainability, that are key for public sector organisations using AI tools and services.
    
3. AI could lead to a long-term skills deficit in key roles where the use of AI prevents the workforce from the practical application of elements of their role. The use of AI tools should not be at the expense of necessary continued professional development. 
    
4. There are concerns AI could impact on workers’ rights - potentially unfairly - and amplify discriminations or bias, if the technologies were used to automate decision making around individuals (such as job applications), or for surveillance and monitoring in the workplace. In terms of the latter, there are concerns about the potential use of automated data to monitor productivity or attendance. 
    
5. Clear guidance and understanding are needed for employees on how to appropriately use AI tools which are available in the context of the official and sensitive information they are dealing with. This will help protect employees from inadvertently using these tools in inappropriate ways, due to lack of training or information. 
    
6. There may be a disproportionate impact on the IT and digital staff within the workforce who will be expected to deliver and embed AI solutions without additional capacity or resource, and with the necessary conditions in terms of expertise, data and culture not being embedded within the organisations. 
    
7. If employers implement AI solutions without the principle of “Human in the Loop” (HITL), errors could be systemised leading to adverse outcomes for service users/employees.
    
8. All these risks are exacerbated if employers fail to engage with employees around the introduction of AI technologies in the workplace. By doing so, they can ensure employees have a say in how the technologies are deployed, can highlight the impact on the workforce and how risks might be mitigated.
    
9. On the other hand, AI offers significant productivity benefits which, if used appropriately, could improve the quality of working life and the well-being of the workforce, as well as improving outcomes for citizens by reducing manual processes and opportunity for human error. 
    
10. By streamlining and automating some elements of service provision, we may be able to reduce demand on stretched public services and increase public satisfaction with how those services can be accessed. 
     
11. AI can support the fight against public sector fraud, including using pattern detection to identify potential irregularities alongside manual work and providing greater assurance for employees.
     
12. There are potential benefits and disbenefits to the bilingual workplace. If the technology is fit for purpose, it may promote true bilingual working through, for example, automated translation in meetings or supporting high quality document translation. 
     
13. However, if models are not built specifically for the Welsh language, they may cause significant inaccuracy and friction in the use of the language. Consideration will need to be paid to the wider implications of automation on the translation workforce.
     
14. There is the potential for AI to support improved quality of working life and workforce wellbeing alongside initiatives like the shorter working week.
     
15. There are potential equality and diversity benefits to AI through providing tools which support accessibility and tailored working practices. For example, AI could be used to support tailored L&D approaches or support neurodivergent employees in day-to-day tasks. 
     
16. To fully realise the potential benefits and guard against the risks of AI, there is a need for a clear focus on the responsible, safe, and ethical use of AI. This requires avoiding several pitfalls that will negatively affect workers and quality of service. 
     
17. There is a risk system which are being marketed as highly performant AI but fail to deliver on that promise. Public sector employees would then have to compensate for those defects. 
     
18. AI systems may be introduced as off-the-shelf products, without the necessary adaptation to the local context before they are put in operation. This can lead to extra burdens on workers, who realise the system does not respond to their needs, or who have to create elaborate workarounds to make everything work – despite the system.   
     
19. The inclusion of an ongoing appraisal and review process would recognise the changing landscape of technological advancements and would ensure the benefits are realised and the threats minimised. 

The objectives of the WPC AI working group include:

1. Developing best practice AI principles, including ‘what not to do’, in the context of the workforce implications.
2. Guidance on how to engage employees when planning to implement AI technologies.
3. Benchmark current understanding of AI, including basic language and terminology.
4. Agree strategic benefits of adopting AI technologies within the workplace using a social partnership approach.
5. WPC explore commissioning a public sector survey on AI, which may influence some/all the above.
6. Review progress of WPC Agreement  - Partnership and Managing Change: Digital Principles.