The rapid emergence of generative AI requires ‘road rules’ or a use policy to be made clear for employees in any organisation.

Many of the risks of AI may already be covered by existing HR policies. But a specific AI policy will provide needed clarity and help avoid wrong assumptions. Such new policy or approach should also reflect the landscape and culture of your organisation. There are three core stages:

1. Determining your position and philosophy
2. Using this to outline your policy
3. Communicating and implementing the policy. 

Consider your organisational culture. If you have values or behaviours that encourage initiative, curiosity and experimentation, then your AI policy should reflect these. If the environment is more tightly controlled or regulated, then your policy may need to reflect that.

Determine why you want an AI policy. Consider the areas of risk management for the organisation, particularly regarding data security (eg ensuring the organisation’s data is not inadvertently placed into the public sphere), accuracy and quality assurance (eg ensuring employees don’t solely rely on outputs from AI tools, but validate the generated information), control (agreeing what is and isn’t reasonable use), and fairness (eg ensuring that use of tools guards against bias). 

Articulate the benefits that you want to achieve. Is the aim to increase productivity, improve quality, reduce time spent on activities, or possibly to accelerate learning? Articulating desired benefits guides usage and reduces fear about elimination of jobs.

Understand current use in your organisation. This may be difficult if there is low trust at present. People who are already using AI may worry they will get into trouble. But it is important to get at least some sense of where, when and how people are using the tools.
  
Consider where AI may be deployed in different roles. Some of this will be driven by the approach of senior leaders but will also be influenced by the nature of role. AI tools can support research, planning documents, writing business papers and policies, assessing data for errors, writing code and beyond.

Maintain compliance with obligations and requirements. These can include aspects such as data security and GDPR obligations, local legislation, or requirements from industry regulators. It is also critical to identify how the organisation will guard against algorithm bias. Determine whether there are new legal requirements driven by AI, or whether your existing policies already cover them sufficiently or should be revised. The new policy may need to articulate the connection to the ones already in place, reminding employees of their existing responsibilities.
In the round, these elements of the organisational culture, structure, current usage, along with matters such as legislative requirements will guide the development of your policy. 

Explain the need for policy. This is a combination of enabling people and providing guidelines. With such a wide range of potential uses, helping people make good decisions is critical.

Open the policy with a statement of how you expect people to behave. This should be grounded in your culture, values, or behavioural expectations. If you seek to be empowering, then emphasise the need for judgement and accountability with AI use. Highlight that AI is a tool – a very sophisticated tool, but a tool nonetheless. Individuals remain responsible for their own work and for the outputs they deliver. For example:

“All employees are expected to apply our core values when accessing, interpreting, and acting on AI-generated outputs. We are all responsible for our own work and ensuring that the use of AI aligns with our mission, values, and behavioural expectations.”

Be clear that AI is a tool, not a replacement for human decision-making. For example:

“AI should only be used as a tool to aid judgement, not replace it. All employees are expected to use their expertise, judgement and empathy when using AI tools. We all have a duty to guard against inherent biases when using AI.”

Emphasise transparency. It’s vital to be clear where AI is being used and where it has or needs to stop. For example:

“We expect all employees to be transparent about their use of AI. Explaining where AI has been used, where it has been tested and with what data inputs, how outputs are validated. This enables us to make the best possible decisions, and to continue to build our expertise and confidence in the use of AI tools.”

Fairness and bias. Recognise machine limitations and imperfections. AI is trained on information it is given, and users should be warned of the risk of bias being fed in and propagated, and should be taking measures to mitigate that risk. For example:

“We are committed as an organisation to mitigate the risk of bias in datasets or AI algorithms. Employees are expected to review and evaluate their models regularly to minimise risk of discrimination.”

Data protection. Outline connection with relevant data protection requirements, and ensure users do not share confidential information or hand over ownership of data in breach of those obligations or safeguards inadvertently. For example:

“Refer to our existing policies in relation to data use and protection. These policies remain in place when using personal AI tools. Employees must treat all data with the same level of security and care and should not place confidential organisational data into open-source tools. In addition, employees must ensure they do not pass ownership of proprietary data to organisations that own the AI tools. Take all due precaution when agreeing to external terms and conditions to avoid inadvertently transferring ownership.”

Management. People managers are accountable for the performance and conduct of their team members, and this remains. For example:

“Our managers are critical to the responsible use of AI tools in the organisation. We expect managers to encourage and discuss with their teams on how to improve their productivity and performance through use. Managers should continue to give coaching and guidance as appropriate.”

Improper use. Be clear on how improper use will be dealt with. For example:

“Where an employee’s action contravenes this policy, this will be managed through our existing performance and conduct policies”.

Ongoing iteration. Acknowledge that AI will continue to evolve. Encourage employees to stay up to date, but be clear on responsibility. Your organisation may want to create a space such as an intranet page or social channel where the organisation and employees can share information, good practice, updates and so on. For example:

“The use of AI is evolving rapidly, and this policy will be reviewed and refreshed regularly. All employees are encouraged to maintain their skills and capabilities in relation to the use of AI by staying up to date on advancements, sharing good practice, experimenting responsibly, and participating in learning exercises.” 

Security, monitoring and audit. Explain the organisational responsibility to maintain its reputation and retain customer confidence, as well as how this will be done. For example:

“The organisation is committed to operating safely and securely. We will measure use, sample activity and audit use of AI, in line with existing data security practices.”

Raising concerns. Be clear on how employees can raise concerns with AI use. This is likely to link to existing policies. For example:

“Where employees become concerned that AI is being used improperly, is driving inequity or bad judgement, or is being used with ill-intent, they should speak up. Employees can raise concerns in the same way as for other issues – directly with the relevant line manager, through grievance or whistleblowing procedures.”

Conclusion. Reinforce your overall approach and philosophy. For example:

“This policy sets out the expectations for our employees in the use of AI. These guidelines and principles will ensure that we use AI to work more effectively and productively. They also ensure that we maintain high quality judgement, creating a culture where we use AI tools to support high quality professional expertise.” 

Consider methods for communicating the policy. Introducing an AI use policy is likely to be sensitive. Employees who are already using AI tools may feel restrained. Others who are suspicious of the tools may feel alarmed. Managers and senior leaders may feel a loss of control.

As such, the policy should go through wide consultation before introduction – share your intention, seek to reassure, and ideally conduct focus groups or surveys to gather input. This will bring in insight as well as an exercise of employee voice, but also prepare the organisation for introduction.

Consider a method that fits with your culture, for example, if it is a reasonably straightforward culture, then you may simply release the policy with a news announcement. If your culture invites experimentation, then you may want to create challenges (eg encouraging all employees to experiment and share their feedback). Videos, newsletters, messages from senior leaders (possibly generated using AI) can all help to encourage use.

Look at opportunities for training, trialling and guiding good use.
  
The broad access to generative AI tools means this can be a free-for-all. Employees will not necessarily understand the ethical risks associated with the adoption of tools. Training should made available to cover:

1. Basic guidance relating to the use of generative AI – getting started, how to ask good prompt questions, things to look out for
2. Ethical guidance (likely to be mandatory) that helps employees understand ethical requirements, the need for compliance with policies, responsible usage
3. How to contribute feedback for development of practice
4. How to speak up when they see things aren’t going well (see ‘Raising concerns’ above).

Monitoring and measurement. The risks of AI use must be managed, but should be balanced with employee privacy. This is likely to require working with wider teams in the business (eg risk and audit functions) to build considerations of ethics and bias safeguards into their review processes. This is to ensure that policies and practices remain fair and don’t discriminate against any set of stakeholders.  

As AI use is potentially more difficult to review or audit, organisations may decide that employees report on AI use as part of their performance management or other practices, or be available for interview/sampling to support auditing. 

In some cases it may be useful to create controls and permissions for AI tools. However, for end-user tools, this may have limited impact as employees could access tools via personal devices.

Maintenance of policy. Set out a framework for review. In many cases, employment policies are reviewed on a three-year cycle or similar. AI will continue to evolve rapidly, with the need to review policy and philosophy more frequently. An annual review cycle may be more suitable initially.

AI tools can be used to support almost every aspect of HR practice. In some cases (eg reward, case management) data may be too sensitive for the use of open-source tools. In other HR functions, there is ample opportunity for use.

There’s also a major philosophical consideration for HR – how do we ensure we access the productivity benefits of AI? For many workers, AI can deliver hours of work in seconds. It can eliminate the need for certain activities, amplify other work, and create new roles.  

HR functions should consider the use of AI in three broad areas:

• how it can reshape HR or people strategy
• how it creates a new paradigm for HR teams
• how it can impact every stage of the employee lifecycle.

AI presents the greatest opportunity to raise productivity for employees who use technology. This creates a raft of considerations for people strategy outcomes.

AI use becomes a new factor to consider alongside other fast-emerging people trends such as positive psychology and psychological safety, risk of burnout, EDI, future skills, good work, four-day week, hybrid working, and others.  

HR strategy needs to consider the impact of AI on all of these elements. It also needs to consider who benefits from productivity gains – does it mean organisations do even more work, or could it create the opportunity for much more flexibility in work, for example, a four-day week.

HR teams often identify developments by benchmarking against other organisations or competitors. AI challenges this model – there are no benchmarks for the use of personal AI. There are no templates for how to adopt use into an HR function. HR teams need to find new ways to develop strategy and working practice.

Create space to try it out. Depending on their workloads and level of comfort with technology, some HR professionals may intentionally avoid using AI. But there’s no substitute for trying it. 

Find opportunities for your team to have time and space to experiment. For instance, conduct an offsite and give them a set of challenges to use personal AI tools to, say shape policies, coach individuals, design training programmes, plan an organisational change. Getting teams to try it in a safe environment creates valuable experience in a short period of time.

Make it part of your HR plan. Ask every part of the HR function to shape a plan on how they would use generative AI in their teams. If they are stuck, encourage them to use an AI tool as a start. Set the tone that teams can uncover more productivity by using the tools.

Get systematic. At present, much AI use is opportunistic. It’s driven by curious individuals applying tools themselves. That might mean it isn’t being used to the greatest effect. One option would be to create a team in HR to shape the plan for testing out AI in HR. That plan should have:

• a framework (eg HR structure or employee lifecycle)
• an approach (eg what experiments to use)
• a means of overcoming barriers (eg upskilling)
• a means of prioritising areas (eg which team first), and
• the opportunity to work with teams.

Understand the gains. Across each part of HR, assess the work that can be accelerated. 

Agree the next steps. As you identify productivity gains, think through the consequences and the opportunities.

You may take a view that in the short term, people can experiment, learn and secure productivity benefits within teams.

For the medium term there are some material decisions that you’ll need to consider as HR leaders, for example:

• Productivity creates capacity. What do we do with that capacity? More of the same, or something different? Does it create opportunity for more development, different work–life arrangements?
• Are productivity benefits felt evenly across the team? Is there a need for organisational redesign in HR?
• What is the impact on management and decision-makers – is there a risk of creating decision-making bottlenecks?
• Impact on stakeholders – what is the capacity of the business to absorb more from HR? 

The CIPD guide to responsible investment in technology can help explore some of these people impact considerations.

Use the approach outlined above to explore opportunities for applying AI against the employee lifecycle and to prioritise actions.

Attraction

Each element of attraction could be enhanced with generative AI, including:   

• Finding routes to candidates, such as identifying the most effective job boards for specific roles, or the events and approaches that can broaden attraction.
• Where social media is used to promote employer brand, generative AI can help shape a content strategy and individual post content.
• Transforming detailed role profiles into candidate-centric job adverts. They can improve inclusive language, brand consistency and clarity.

Selection

Generative AI tools can help candidates be much better prepared for job search – it can help research companies, prepare presentations and prepare answers to questions. 

This means that old selection practices may need re-examining and updating, including:

• Consider the questions you ask, and how you score them. There may be a need to weight specificity in answers over generalisations. AI can help people to answer thematic questions, but less so about specific experiences.
• Consider the use of psychometrics. These provide an additional lens on candidates, giving an opportunity to assess candidates’ attitudes and mentality, and there’re no right or wrong answers.
• Make ability to use AI a selection criteria – this is effectively a demonstration of curiosity and initiative. To test for this, ask candidates to prepare a case study, and specifically ask them to use AI to help them. This gives the opportunity to ask questions about the subject, but also to ask how the candidate used AI, where they added to it, and how they applied judgement.

AI tools can provide a useful service in sifting and ranking candidates based on their written applications. However, there are some material risks that you should consider:

• If you are using an open-source tool, how do you protect candidate data?
• How do you minimise risk of discrimination (eg make sure you are clear on the selection criteria being applied, and how the AI has been trained)?
• How are you tracking and correcting any potential discrimination from the AI tools?

These risks are mitigated by HR professionals investing the time to properly set up AI tools – being clear on criteria, providing quality context, testing with sample data and refining.  Treat AI outputs as draft recommendations – always bring human expertise and judgement to bear in making decisions. AI use prepared in this way can guard against human bias, and maintain consistency in assessing applications (AI does not get bored or tired). It is critical that the professionals own the outcomes, and actively test to mitigate bias.

Induction

Induction processes can be difficult to shape and maintain. AI can support with planning induction programmes, stakeholder meetings, sharing relevant information, phasing training and reading for new starters. 
 
AI-powered search engines and chatbots could change this process, creating a much more interactive way of getting to know the company, or scouring organisational resources to accelerate learning.

Onboarding

Generative AI tools could be used to create more personalised onboarding plans based on the role, the new hire’s experience, and interview outputs.
  
Proprietary AI tools such as Microsoft Copilot can be used to provide 24/7 support to new hires to answer common questions (removing reliance on the old HR FAQ documents). They could help shape personalised learning plans for the new hire based on assessment of interview notes, experience, and the nature of the role. They can provide support for hiring managers in clarifying important tasks, and significantly reduce the time spent searching HR materials or speaking to HR teams.

Compliance training

Compliance learning can be difficult to design and to maintain, as organisational policy or external regulation develops. For policy owners, AI tools could accelerate this activity through review of current copy and identifying legislation changes.

AI can test to ensure that learning modules prioritise the key learning points for users, and can be used to write tests that reflect the most critical learning points.

Existing approaches risk becoming less effective as employees click-through videos and slides then use AI or simply search engines to answer tests.  

There is an opportunity to rethink the approach to compliance learning overall. For instance, AI tools could be used for open-text questions, replacing the usual multiple-choice tests. There is a greater degree of rigour in assessment and more self-driven learning.

A further step would be to shift from the current learning modules to employees sourcing information for themselves, using AI tools. This could be developed into coaching style conversations with chatbots to explore the area of learning.

Organisational policies

It is possible to use an AI tool to draft a policy on a specific subject incorporating previous policy, updated expectations, applying a house style or similar. This can accelerate the development of policies, and create consistency in style of policy. However, it remains important for the professionals to review and own the content.

Where internal tools are available, an AI-powered chatbot can point employees to relevant sections of policy as and when required. An alternative approach is to use AI to produce user guides, or process walkthroughs. This is particularly helpful if policies are reflecting complex legislation. In areas of high risk or ambiguity, it would be advisable to have employees check with the HR team. 

Reward

Reward practice can benefit from AI tools, but is an area where data security must be considered. It is worth teams exploring where data can be secured (eg by ensuring employees cannot be identified) or identifying tools that can be used in-house to manage company data.

One simple action to manage security is to use extracts of data, with unique identifiers in place to replace identifiers such as name or employment number.
  
With measures in place, there are opportunities to use AI for activity such as:

• Analysing data including performance metrics, productivity, skills and activity, to provide more thorough benchmarking for reward decisions.  
• Analysing internal and external reward information to ensure fairness and alignment. This has the potential to identify discrepancies and trends over time which in turn can help address pay inequity.
• There is potential for employees to use generative AI to really personalise their use of rewards and benefits, for example ‘discussing’ their personal position and the outline of company benefits with AI. This could be done with personal tools, but could become more targeted if enterprise tools were adopted.
• Reward teams can potentially identify how benefit packages are used, what is the relative value of different elements, and produce targeted communications to help colleagues make informed choices.
• There’s also the potential to use AI to improve the speed and quality of pay benchmarking submissions.

Ultimately, there may be enterprise-level, data-compliant AI tools that are better suited to help in this area, but personal tools can also be of good use.

Wellbeing

As wellbeing offerings are extended by the organisation, generative AI can help navigate the offering. For example, AI can help wellbeing professionals to access information, accelerate learning, and build strategy.

For employees, AI tools could help navigate resources to find what is most suitable. There is also great potential for coaching – employees and managers can use generative AI to help frame conversations, gather information, identify good practice and get to more personalised solutions faster.

Learning and development

AI can accelerate the development and design of learning programmes, creating outlines,  detailed content, exercises and notes for facilitators. Company behaviours or priorities can be updated, allowing developers to ensure programmes remain current.

There’s an opportunity to encourage learners to use AI as part of their pre-reading, or as follow-up exercises. This gives the opportunity to build confidence in AI use as well as the subject of the course.

Finally, it is common for organisations to have catalogues of learning resources. AI could be used to:

• Help employees find and summarise content from the library (less time searching, more time learning)
• Remove the need for some of these catalogues altogether. AI tools can be used to find and identify openly available material, summarise it and give core information. There is a risk regarding the accuracy and variety of source content, and qualifying parameters should be established. Shaping guidance on using AI for learning prompt good questions of how to connect it to role needs and what are important roles for development teams.

Management

One way AI can support managers is to help them identify and manage the demands on their time. But as generative AI is embedded into software packages (eg Microsoft Copilot), it may be more about using enterprise-level rather than personal AI to schedule 1-1s, team meetings, capturing action points and so on.

Another use is to help managers understand what policies apply in specific situations, and how to apply them. AI tools can help managers to boil policies down into a short list of tasks they need to complete, give them guidance on how to approach conversations with colleagues. This significantly simplifies HR advice, allowing HR professionals to provide more in-depth support for complex situations.

More importantly, management skills change with AI. As employees use AI more, managers need to have quality coaching skills, the ability to assess and understand the use of AI. As team members shape pieces of work, managers will need to prepare questions such as:

• “How did you use AI to help with this piece of work?”
• “How did you validate the inputs?”
• “What other sources did you use?”
• “How did you bring organisational context to your work?”
• “Where did you add your judgement?”
• “How did you ensure data security?”  

There is also a wider consideration. In recent years demands on managers have increased, as issues like worker wellbeing, inclusive culture and managing hybrid teams or flexible working all compete for attention. AI use adds another layer as it requires managers to be able to coach others around AI use. HR teams should consider managers’ capabilities and the impact on their capacity. This may require some role and/or organisational design changes. 

Performance management

AI tools can support each step of the performance management journey, reducing the administrative burden to create more focus on quality management.

Tools can help craft objectives (using historic ones), organisational priorities or strategies, assess existing workload and external data. Cascading objectives can similarly be accelerated. All this activity can increase organisational alignment and improve collective performance.
 
Development plans can be developed in a similar way – individuals can receive guidance from AI tools on how to identify development priorities, how to work towards them, along with specific suggestions. Increasing the quality of development plans and the speed of generation allows more time for thought.

Enterprise-level AI tools magnify the potential to assess what people have done. AI and workplace analytics embedded into applications like Microsoft VIVA can identify where people have spent their time, who they’ve spent time with, what they’ve prioritised and what they’ve neglected. It can allow for much more evidence-based performance discussions.

A key element of performance management is ongoing feedback. Many people will spend hours crafting feedback to ensure their meaning is clear, or to avoid offence. AI can help in crafting and editing messages and adopting the right tone of voice.

Career progression 

AI tools can help employees identify their own career paths and opportunities based on their experience and preferences. If an employee has a known goal (eg chief people officer), information on their starting point (eg reward analyst job description and their CV), and potential steps to get there (eg competency frameworks, HR team structure, CIPD Profession Map), then AI can trawl all these sources to recommend a route.

Frequently all these information points will exist within organisations, but bringing them together manually would be time-consuming.

Talent assessment and profiles

Talent processes can be sensitive, and there are risks relating to bias in the dataset due to historic data (eg senior leaders are potentially more likely to be male, white, straight, university educated).

Nevertheless, AI can be used to help collate, analyse and plan talent outcomes. It can be used to generate standard talent profiles based on data such as CVs and LinkedIn profiles and use that data to identify cross-functional succession candidates, suitable people for key projects and longlists for internal vacancies.

As already mentioned, the requirement for human judgement and validation is critical here to avoid bias in the system.

Guidance

AI tools can make this significantly simpler. Applying a generative AI tool to internal policies would allow managers and employees to ask for information, for guidance on how to handle a situation, for the right forms to complete, to check for deadlines and so on. It can shift to a true self-service model, based on questions that managers and employees really do ask, rather that HR teams having to develop sets of FAQs. This is likely to be significantly easier and more tailored as enterprise-level tools incorporating generative AI become more widely available. 

This then lifts HR teams into a more value-adding role, with more focus on advice, quality assurance and intervention in complex situations.

There is a huge shift for HR professionals in terms of skills and mindset. There will be concern from employees that they will lose their role if AI can do it in a fraction of the time. Similarly, there will be concern about changes in ways of working or skillsets for some members of the HR team themselves.  

There may be the need for a skills audit, assessment of career goals, or cultural piece of work to encourage HR colleagues to seize the opportunity.

There is a tension for HR teams. A great deal of HR practice is confidential and sensitive. Some HR activity is likely to need internal (enterprise-level) AI tools to be deployed to ensure data security. However, even in areas that deal with sensitive data, there may be ways to anonymise data to allow use of open access, personal AI tools.

At the same time, there is a lot of HR work that is not proprietary or sensitive. In these spaces, using open access AI becomes entirely reasonable, meaning there are no real barriers to use today.

HR teams should consider the case for adopting generative AI technology ‘officially’ through enterprise-level tools, whether that is replacement of existing applications, saving managerial time, or improving access to information for managers and employees.

People in your organisation, those working with you, and those applying to your organisation will be using AI tools. It is changing the way we work, so HR teams need to get and stay ahead of it. Equally, it is critical to remember that AI is a tool and only part of any solution. AI may get you two-thirds of the way there, but the final and crucial third is human judgement, expertise and application.

Sometimes HR can be cast in the role of the police or controller. For instance, writing and enforcing social media policies might have been a relatable experience for many a practitioner. Or maybe more recently this would have been for use of Teams or Zoom. The sheer variety of AI use makes a hard, rules-bound approach almost impossible. Instead, we should lean into the possibility, use high-quality management and human judgement, guided by core values, to lead the way on application in this brave new world.

This guide was written on behalf of the CIPD by Francis Lake, founder of Green Juniper, a people advisory consultancy.