How people professionals can develop, deploy and use AI in an ethical, legal and sustainable way

AI governance establishes the controls needed to develop, deploy and use AI ethically, legally and sustainably. Much like building cars with seatbelts and enforcing road rules so that people can get to their destinations safely and quickly.  

AI governance mirrors corporate governance by establishing accountabilities, processes and systems for organisational success. Good AI governance supports overall good corporate governance.

AI governance is necessary for effective AI use because it provides the controls to balance the opportunities and risks of AI. To drive performance with AI sustainably, organisations must use AI responsibly or risk losing valuable talent, facing litigation, and damaging brand and reputation. This means that ethical use of AI should be seen as foundational to an organisation’s stability.

People professionals should be involved in proactively shaping governance particularly where AI impacts people. This includes safeguarding employees and their data and addressing ethical and legal people risks, and preparing organisations for the impact of AI on roles and careers.  

Wherever AI has a human impact, the people team needs to be just as involved as IT, legal and compliance colleagues. AI must fit with the people, culture, structure and tasks for an organisation to achieve the desired returns on investment.  This is also true of rolling out any change management initiative in an organisation. AI is only one element, and the rest is about people and how they work together. 

The people team needs to partner with IT, innovation, legal and compliance teams to deliver the change. The role that the people team plays will depend on what’s being deployed. For example, deploying an AI customer service chatbot will likely involve the people team steering communication, training and consultation activities. On the other hand, deploying an AI-driven recruitment system will entail everything from drawing up a request for proposal (RFP) to monitoring and auditing the system when it’s in live operation. 

Your knowledge of AI governance, the law and your organisation as well as your professional expertise will help you discuss the impacts of AI in an organisation with credibility. Responsible AI governance should adhere to AI principles as well as general and industry specific regulations. Some international standards underpin national legislation. As AI is an emerging field, existing frameworks are being reviewed and new ones developed. You may need to help your colleagues understand how these interact or identify how changes will impact governance. 

Ways in which people professionals can build AI governance knowledge include: 

• Gain a basic understanding of how AI works. For people professionals who are eager to learn more about HR’s role in this area, exploring CIPD AI resources and the CIPD Learning hub are good starting points 
• Know what people-centred AI principles typically cover. OECD’s AI principles, for example, call for inclusive growth and sustainable development, fairness and privacy, transparency and explainability, robustness, security and safety, and human accountability. It’s signed by over 48 countries. UNESCO has similar AI principles, which applies to 194 member states. 
• Be aware of key standards, regulations and guidance that supports people centred AI principles. For example, the CIPD guide produced with the British Standards Institution on navigating AI standards goes over key standards in this area. Meanwhile, familiarising yourself with the EU AI Act is a good idea, as this applies to organisations with employees and customers in the EU. UK employers should also make sure they're familiar with other laws and guidance that govern technology use at work, such as GDPR 2018 and the Equality Act 2010 as well as industry specific regulations and guidance, where applicable. 
• Discuss with peers in the CIPD Community. This can be done by posting and replying to questions in our AI in the workplace forum. 
• Mentor IT or AI professionals. If there aren’t any in your organisation, you could look for opportunities to mentor IT or AI professionals in other organisations.  
   

Striving for good data quality is a basis for building good AI systems. AI can process and synthesise large volumes of data quickly. Using AI to process incomplete and inaccurate data risks churning wrong outputs on a large scale.  

Ways you and your team can build people’s trust in data and lay the foundation for responsible AI governance include:  

• Understand data protection and privacy principles. Ensuring your team understands the data protection and privacy principles in the countries where your organisation operates is key.  
• Understand data provenance, data access and data use. It’s important to ensure your team understands where the data has come from, who can access the data and how it’s used. 
• Formalise data governance. Rules such as which systems can talk to each other and who can see what data should be built into data policies, systems maintenance processes and approval workflows. A people systems lead responsible for managing people data could support in this area. 
• Manage data expectations. Unlike financial data, employee data usually needs consent from the individual to use in a specific way. This means you will have to manage colleagues’ expectations, so they know why workforce data is rarely complete and where the gaps are. The CIPD’s people analytics guide discusses the data quality conundrum.  
• Interpret data wisely. Demonstrating sound judgment when interpreting workforce data of varying quality fosters transparency. For example, if only half of employees responded to an organisation-wide employee survey then admit that the findings might not represent general opinion. Find other ways to check and reveal insights not flagged in the survey data, like speaking with employee groups whose teams have low survey response rates. 
• Work with colleagues to improve data quality. Working with colleagues to improve the quality of employee data so that it’s good enough for the intended purpose is essential. Suppose the headcount data in management reports are inaccurate. The figures don’t tally between systems. To solve this, agree how to calculate headcount. Then, to simplify reporting, look at whether headcount data collection could be automated across different systems.  

Effective AI governance relies on considering multiple specialist views for each AI system. People professionals should role model auditing AI governance effectively and apply people expertise.  

As people professionals, you can support organisational strategy by advocating for meaningful ways to oversee AI. You are well placed to foresee the people impact of AI and the potential business impact if humans are not adequately involved. Common problems include: 

• The potential for AI outputs to be biased against specific groups. 
• The impact on job design which may trigger consultation obligations or fears regarding job security. 
• The risk that human oversight in AI-driven activities is not meaningful or effective. 

Do explore possible second and third order effects of an AI system. For example, making use of an AI system optional could create performance gaps if certain age or gender groups use the system more than others. Good and accessible training can solve this provided people have the time and the goals are achievable. 

You should encourage exploration with what-if scenarios to help colleagues understand where your organisation’s AI governance could be improved, such as: 

• Whether an impact assessment process is in place before the AI system is deployed. 
• Whether capability gaps have been identified and if there are development plans to address those gaps (eg AI skills, foundational skills). If there’s less need for certain skills, then discuss what your organisation will do.  
• Whether the right people have been assigned to monitor the performance of and solve problems arising from an AI system.  
• Whether there’s sufficient transparency relating to the use of AI tools.  
• Whether it’s clear how to report and escalate problems. And whether you have the right senior people overseeing this who can advise and step in when needed. 

To learn more about implementing good practice AI governance, read our guide on navigating AI standards produced with the help of the British Standards Institution. 

The people team should be involved in the organisation’s AI governance discussions given the wide-ranging impact of AI on people. Regulatory compliance in AI governance is, however, typically in the remit of the legal, risk or compliance teams in large organisations. Colleagues outside these teams might not feel confident to challenge. Nor might they know how to deal with ethical issues like inclusion, fairness and safety.  

You can lean on your existing knowledge of people, employment law and experience of developing policies to discuss AI governance pragmatically. By being aware of the letter and spirit of employment-related AI regulation, you can discuss pragmatic ways to comply. For example, the EU AI Act flags AI use in employment as ‘high risk’. However, by meeting certain criteria some high-risk use, such as shortlisting candidates, can be downgraded to limited risk. The EU AI Act compliance checker questionnaire is a useful tool for checking risk level.  

Discussing AI governance also gives you the chance to engage with colleagues about your organisation’s approach to ethics in general. Suppose you’re discussing how to avoid discriminatory hiring in AI recruitment systems. First, you need to agree collectively what discriminatory hiring means and establish what is important to your organisation. Is the priority simply to avoid unlawful direct and indirect discrimination? Or does your organisation also want to create a more diverse workforce? Is it about legal compliance, business benefit or organisational values? Only when you’ve established the priorities can you discuss practical ways to avoid discriminatory hiring. 

Aligning leadership behaviours, policies and procedures, and incentives ensures that people-centred AI principles become part of the organisation’s culture. Creating the conditions for all employees to practise people-centred AI principles at work helps normalise good AI. Because while people professionals have a key role in AI governance, it’s unlikely you or your team can join every discussion. 

Making sure employees know how the AI systems they use work is an example of supporting the AI principles of transparency and explainability. Let’s say a people team is launching an AI wellbeing app for employees to get personalised health nudges. Do tell employees what data the app monitors, why it’s useful for them and how it benefits the organisation, and whether they can opt out of using the app. Providing employees with FAQs can help allay concerns employees may have about the organisation’s use of their health data. 

Ways you can embed AI-centred principles include: 

• Instil a people-centred AI principles mindset. Use your organisational development skills to instil a mindset where employees practise people-centred AI principles, such as accountability, transparency, explainability, and inclusive growth. You can role model responsible AI governance when deploying AI features in people systems, and coach leaders on how they talk about experimenting and using AI at work. 
• Introduce an AI policy or charter. Introducing an AI use policy sets the tone and guardrails for AI use. Many organisations are adopting an AI charter that support people-centred AI principles.     
• Agree on a framework for working together if you have trade unions. If your organisation has trade unions or employee representatives, it’s a good idea to agree on a framework for working together. Some trade unions may have their own guidelines you can use. Involving employees is vital for any change initiative. Talking with trade unions and employee representatives is another way to build trust and deal with worries about job security. 
• Implement processes that support responsible AI governance. 
  Implementing processes such as the AI standards covered in the CIPD’s guide to AI standards helps build accountability in managing an AI system. You can help amplify accountability by fostering an organisational climate where people feel safe to speak up and take responsibility for solving problems. 
• Recognise and reward behaviours that support people-centred AI principles. If your organisation holds innovation competitions, ensure employees get recognised and rewarded for innovation grounded in people-centred AI principles.