Artificial Idea | AI careers · practical prompts · no hype Thursday, October 30, 2025 · Issue #26 · Prompt Tutorial

The HR playbook

5 prompts for HR professionals to handle the questions AI can't answer

AI is changing every function in every organisation. HR is changing faster than most, and the professionals inside it who understand what AI can and cannot do are the ones being asked to lead the transition.

Issue #25 made the case that the bifurcation in the labour market is real, measurable, and compounding at a rate that makes engagement with it increasingly urgent. This issue turns to a function that sits at the centre of that bifurcation in every organisation: human resources.

HR is in an unusual position in the AI transition. It is simultaneously one of the functions most affected by AI adoption, responsible for managing the workforce implications of a technology that is restructuring the workforce, and one of the functions where the limits of AI are most consequential. The decisions HR professionals make about people, about fairness, about development, and about organisational culture cannot be delegated to a tool without losing something essential in the delegation.

The HR professionals navigating this well are those who have developed a precise understanding of where AI adds genuine value in their function and where it does not, and who have built workflows that use AI for the former while protecting the human judgment required for the latter. These five prompts are designed for that purpose.

They are also relevant to any professional who needs to understand how HR actually works, because the professionals who understand the function that manages their career trajectory are consistently better positioned than those who treat it as an administrative background process.

Prompt 1: The job architecture analyser

The problem it solves: evaluating whether a role's current job description accurately reflects what the organisation actually needs, which in most organisations it does not, because job descriptions are rarely updated with the frequency that actual role requirements change.

You are a senior HR business partner with 
deep experience in job architecture and 
role design helping me evaluate whether 
a current job description accurately 
reflects the role's actual requirements.

Current job description: [paste]

What the role actually involves based 
on my knowledge of the function and 
the organisation's current priorities: 
[describe in as much detail as possible]

The organisational context: [what is 
changing in this function, what pressures 
the team is under, what the organisation 
is trying to achieve in the next 
twelve to eighteen months]

Please:

1. Identify the gaps between what the 
   job description specifies and what 
   the role actually requires, in order 
   of significance
2. Identify any requirements in the 
   job description that are no longer 
   relevant to what the role actually does
3. Assess whether the current job description 
   would attract the profile of candidate 
   most likely to succeed in the role 
   as it actually exists, and where it 
   would attract the wrong profile
4. Draft a revised role summary of 
   three to four sentences that accurately 
   reflects what the role requires and 
   what kind of professional would 
   thrive in it
5. Identify the two or three capabilities 
   that are currently missing from the 
   job description that AI-augmented 
   work in this function now requires

Flag any area where the gap between 
the described role and the actual role 
is large enough to create a legal or 
equity risk in the hiring process.

The legal and equity risk flag is not incidental. Job descriptions that do not accurately reflect role requirements create measurable risks in hiring processes, particularly when the gap between the description and the actual role affects which candidates self-select into the pipeline. HR professionals who conduct this analysis proactively are managing a risk that their organisations often do not know they are carrying.

Prompt 2: The difficult conversation preparer

The problem it solves: preparing for the specific category of HR conversation that is most consequential and most difficult, which is the conversation with an employee whose performance, behaviour, or situation requires a direct, honest, and legally defensible discussion.

You are an experienced HR director helping 
me prepare for a difficult employee conversation.

The situation: [describe the performance issue, 
behavioural concern, or circumstance requiring 
the conversation, including the history, 
what has been communicated previously, 
and what outcome the conversation needs 
to achieve]

The employee context: [their tenure, 
their general performance history, 
any relevant personal circumstances 
you are aware of, and your assessment 
of how they are likely to respond]

Organisational context: [any relevant 
policies, previous commitments made, 
or stakeholders who need to be 
kept informed]

Please help me prepare:

1. The opening statement: how to begin 
   the conversation in a way that is 
   direct without being aggressive, 
   and that establishes the purpose 
   clearly without creating immediate defensiveness
2. The core message: the three things 
   that must be communicated in this 
   conversation regardless of how 
   it develops, stated in plain language
3. The likely responses: the three 
   most probable reactions from this 
   employee and how to handle each 
   one without losing the thread 
   of what the conversation needs to achieve
4. The documentation: what should be 
   documented before, during, and 
   after this conversation to create 
   an accurate and legally defensible record
5. The follow-up: what needs to happen 
   within 48 hours of this conversation 
   to ensure the outcome is properly 
   supported and tracked

Constraint: this preparation must be 
grounded in what is fair and honest 
to the employee, not only in what 
is legally defensible for the organisation. 
Both matter. The preparation should 
reflect both.

The constraint that the preparation must reflect what is fair to the employee, not only what is legally defensible for the organisation, is the one that makes this prompt useful for HR professionals who take their function seriously rather than those who treat it as organisational risk management with a human face. The two objectives are compatible in well-run organisations and in tension in poorly run ones. The prompt is designed to serve the former.

Prompt 3: The organisational culture diagnostician

The problem it solves: turning qualitative signals from employee surveys, exit interviews, and informal feedback into a structured diagnostic that identifies the specific cultural dynamics most likely to affect performance, retention, and the organisation's ability to navigate change.

You are an organisational psychologist helping 
me diagnose cultural dynamics from qualitative 
employee data.

Data sources available: [describe the data: 
employee survey results, exit interview themes, 
informal feedback, manager observations, 
or any combination]

Here is the data: [paste or summarise]

Organisational context: [the organisation's 
current situation, any significant changes 
underway, and the specific cultural challenges 
leadership has identified or suspects]

Please:

1. Identify the three most significant 
   cultural dynamics visible in this data, 
   described in terms of the underlying 
   organisational behaviour rather than 
   the surface symptoms
2. Identify any pattern in the data that 
   is likely to be a leading indicator 
   of a problem that has not yet fully 
   manifested, and why
3. Identify any gap between what the 
   organisation believes about its culture 
   and what the data suggests is 
   actually true
4. Assess which of the identified dynamics 
   is most likely to affect the organisation's 
   ability to navigate the specific changes 
   currently underway
5. Propose one intervention for the most 
   significant dynamic that is specific 
   enough to be implemented within 
   thirty days and measurable enough 
   to evaluate within ninety

Flag where the data is insufficient 
to support a confident diagnosis and 
what additional data collection 
would most improve the analysis.

The instruction to identify gaps between what the organisation believes about its culture and what the data suggests is actually true is the most politically sensitive and most analytically valuable component of this prompt. Organisational cultures are frequently misperceived by the people inside them, particularly at senior levels where the filtered nature of upward communication produces a systematically distorted picture. Naming that gap explicitly, with data support, is the diagnostic work that makes cultural interventions possible rather than cosmetic.

Prompt 4: The talent risk identifier

The problem it solves: identifying which employees represent the highest retention risk before they resign, using the patterns in available data to surface leading indicators that are consistently present before voluntary departures but consistently missed until it is too late to address them.

You are a workforce analytics specialist 
helping me identify talent retention risks 
before they become departures.

Available data about my team or organisation: 
[describe what you have access to: 
performance review history, promotion 
timeline data, compensation relative 
to market, engagement survey results, 
tenure patterns, or any other relevant signals]

Here is the data or a summary of it: 
[paste or describe]

Organisational context: [the current 
environment, any recent changes that 
might affect retention, and the roles 
or individuals where departure would 
have the highest organisational impact]

Please:

1. Identify the profile most at risk 
   of voluntary departure based on 
   the patterns in this data, described 
   in terms of the characteristics 
   and circumstances rather than 
   individual names
2. Identify the leading indicators 
   present in this data that research 
   consistently associates with 
   pre-departure behaviour
3. Assess which specific circumstances 
   or changes in the current environment 
   are most likely to trigger departure 
   decisions among the at-risk profile
4. Propose three retention interventions 
   ranked by their likely effectiveness 
   for this specific profile, with 
   a rationale for each
5. Identify the intervention that 
   is most commonly attempted and 
   least effective for this type 
   of retention risk, so it can 
   be deprioritised

Be specific about what the data 
supports and what is inference. 
Retention risk analysis that 
presents inference as certainty 
produces interventions that 
address the wrong problem.

The instruction to identify the commonly attempted but least effective intervention is the component most likely to produce immediate practical value. Organisations facing retention risk tend to reach for the most visible intervention, typically compensation adjustment, regardless of whether compensation is the actual driver of the risk. Identifying the mismatch between the habitual response and the actual problem is often more valuable than identifying the right response, because it prevents wasted resource allocation before the right response is considered.

Prompt 5: The AI policy drafter

The problem it solves: developing a practical, specific, and enforceable organisational policy for AI tool usage that addresses the real risks without being so restrictive that it prevents the productivity gains AI enables.

You are an HR and employment law specialist 
helping me draft an organisational AI 
usage policy that is practical, specific, 
and legally sound.

Our organisation: [size, industry, 
the types of AI tools currently in use 
or under consideration, and the specific 
risks most relevant to our context]

Our current approach: [describe any 
existing guidance, formal or informal, 
about AI tool usage]

The specific concerns we need to address: 
[data privacy, intellectual property, 
accuracy and verification requirements, 
equity in AI-assisted decisions, 
client confidentiality, or any other 
specific concerns relevant to our context]

Please draft a policy framework covering:

1. Permitted and restricted uses: 
   specific enough that an employee 
   knows whether a given use is 
   permitted without having to ask, 
   but not so prescriptive that 
   it prevents uses the organisation 
   would benefit from
2. Data handling requirements: 
   what categories of information 
   should not be input into AI tools, 
   with a clear rationale for each restriction
3. Verification and accountability: 
   the standard of review required 
   for AI-assisted outputs before 
   they are used, communicated externally, 
   or acted upon, calibrated to 
   the risk level of different use cases
4. Disclosure requirements: when and 
   how employees should disclose 
   AI assistance in their work, 
   to internal and external stakeholders
5. Review mechanism: how the policy 
   will be updated as AI tools and 
   their capabilities evolve, 
   with a specific review cadence

Constraint: the policy should be 
readable by a non-specialist employee 
in under ten minutes and actionable 
without legal interpretation. 
Flag any area where legal advice 
specific to our jurisdiction 
is required before finalisation.

The constraint that the policy should be readable and actionable without legal interpretation reflects one of the most consistent failure modes in organisational AI policy development. Policies drafted primarily to manage legal risk produce documents that are comprehensive, defensive, and largely ignored by the employees they are meant to guide, because those employees cannot extract from them a clear answer to the practical question they are actually asking, which is: can I use this tool for this purpose?

A policy that answers that question clearly for the most common use cases, while flagging the edge cases that require further guidance, is followed. A policy that requires legal interpretation for every use case is filed.

A note on what these prompts cannot do

These five prompts cover the analytical and preparatory work in HR functions where AI adds genuine value. They do not cover the work where AI does not belong.

The decision to terminate an employee cannot be delegated to an AI analysis. The judgment about whether an organisational culture is healthy requires human presence and relationship that no data summary can substitute for. The conversation with an employee in distress requires a human being, not a prepared script. The assessment of whether a candidate is right for a role involves dimensions of judgment, intuition, and interpersonal reading that sit outside what these tools can produce.

The HR professionals who use AI most effectively are precise about this boundary. They use the tools to do the analytical groundwork faster and more thoroughly, freeing the time and cognitive capacity for the human work that the analysis is meant to support. The boundary between the two is not always obvious and it shifts as the tools develop. Maintaining clarity about where it currently sits is part of the professional responsibility of anyone using these tools in a function where the consequences of getting it wrong are borne by people rather than processes.

Monday we are examining the most counterintuitive finding to emerge from the AI and careers research this year: the characteristic most correlated with successful navigation of the AI transition is not technical skill, not formal education, and not seniority. It is something simpler, less glamorous, and more within the reach of every professional reading this than any of those variables.

The finding is well-supported and almost entirely absent from the mainstream AI career conversation. Monday's issue puts it where it belongs.

— The Artificial Idea team

Keep Reading