What Workplace AI Ethics Actually Involves
Here's the thing nobody tells you when you start using AI at work: the tools are easy. The ethics are hard. Not because the questions are abstract or philosophical — but because they show up in everyday situations where the "right" answer depends on context, relationships, and judgment calls that no policy can fully anticipate.
Should you disclose that AI helped you write a client proposal? What about a performance review? An email to your team? A job application cover letter? The answer to each of these is different, and the reasoning behind each answer reveals something important about how we think about authorship, honesty, and the value of human effort in an AI-augmented world.
We've wrestled with these questions ourselves — and we've gotten some of them wrong. Early on, we used AI to help draft a training curriculum without disclosing it to the team reviewing it. The content was good, genuinely useful, heavily edited by us. But when a colleague later asked "did you write all of this?" and we hesitated — that moment of hesitation told us something. If disclosure feels uncomfortable, it's usually because the norms haven't caught up with the technology. And when norms are unclear, the people who establish transparent practices early earn trust that compounds over time.
If you've ever used AI for something at work and wondered whether you should mention it — or if you've worried about what AI means for your career, your team's jobs, or your industry's future — those concerns aren't signs of technophobia. They're signs that you're thinking about AI the way a responsible professional should. This path gives you frameworks for the questions that don't have easy answers.
What makes workplace AI ethics genuinely challenging is that reasonable people disagree. There's no universal rulebook. But there are principles that help you navigate grey areas with integrity — and practical frameworks that make the hard calls easier. That's what we're building here.
Get Started
Before we explore frameworks and policies, let's start with your own intuitions. Read each scenario and note your instinctive reaction — then we'll examine why:
- You use AI to help write a report for your manager. Do you mention it?
- A job candidate's cover letter was clearly AI-generated. Does that affect your evaluation?
- Your company uses AI to screen resumes before a human sees them. Are you comfortable with that?
- A colleague presents an AI-generated analysis as their own work in a team meeting. How do you feel?
- You use AI to prepare for a salary negotiation — generating counter-arguments, market data, and persuasion strategies. Is that ethical?
If your answers were all immediate and certain, you might be oversimplifying. If some of them made you pause — especially if your answer to similar-sounding scenarios differed — you're already engaging with the real complexity. Most ethical questions in AI aren't about clear right and wrong; they're about competing values (efficiency vs. transparency, individual advantage vs. collective fairness) that need to be balanced thoughtfully.
Now try this: open your AI tool and ask it, "Is it ethical to use AI to write a cover letter for a job application?" Read its response. Then ask: "Now argue the opposite position." Notice how convincingly it argues both sides. That's not a flaw — it's a reminder that these are genuinely contested questions where thoughtful people land in different places.
Core Skill 1: Attribution and Transparency
The question "should I disclose AI use?" comes up constantly — and the answer is almost never a simple yes or no. It depends on what you created, who you created it for, and what expectations exist in that relationship. A framework helps you navigate this without agonising over every email.
Last verified: March 2026
🔢A majority of knowledge workers now use AI regularly at work
but only a fraction have disclosed it to their managers or clients. The gap isn't usually dishonesty — it's uncertainty about norms. Most people don't hide AI use; they simply don't know when mentioning it is expected, appreciated, or irrelevant.
Source: Microsoft Work Trend Index 2024–2025; Salesforce workforce research
An attribution framework that actually works
Last verified: March 2026
We use a three-tier approach based on the level of AI involvement and the stakes of the output:
⚖️ AI Disclosure Framework: When and How to Attribute
AI Involvement Level Description Disclosure Recommendation Examples AI as assistant AI helped brainstorm, check grammar, suggest edits — you did the thinking Optional. Mention if asked or if norms require it. Using Grammarly, asking AI for synonym suggestions, brainstorming session titles AI as collaborator AI generated significant portions; you edited, curated, directed Recommended. Acknowledge the collaboration. AI-drafted report sections, AI-generated first drafts you heavily revised, AI-structured analysis AI as creator AI produced the core output; your contribution was prompting and selection Required. Non-disclosure risks trust and credibility. AI-written articles published under your name, AI-generated code submitted as your work, AI-created presentations
Context matters more than rules
Last verified: March 2026
The same level of AI involvement might require different disclosure depending on context:
- Academic work — virtually all institutions now require AI disclosure. The trend is toward treating AI assistance like citing a source: not shameful, but must be acknowledged.
- Client deliverables — depends on the client relationship and contract terms. Some clients explicitly prohibit AI use; others expect it. When in doubt, ask.
- Internal communications — team norms vary widely. In some teams, using AI for emails is as unremarkable as using spellcheck. In others, it would feel deceptive.
- Creative work — the highest sensitivity. Readers, audiences, and buyers of creative work often feel deceived by undisclosed AI involvement, even when the output quality is high.
- Job applications — a rapidly evolving area. Most recruiters now assume some AI involvement and care more about whether the candidate can do the work than whether AI helped with the application.
Knowledge Check
You've used AI to draft a detailed client proposal. You then spent two hours editing, reorganising, adding your own analysis, and customising it to the client's specific situation. The final document is about 60% AI-generated structure with 40% your original contribution. The client asks: "Did you write this yourself?"
📈 Disclosure Norms Across Professional Contexts (2025-2026)
Source: Industry estimates based on Reuters Institute research, SHRM hiring surveys, and academic integrity studies
There's a useful test for disclosure decisions: imagine your audience discovers the AI involvement six months from now. Would they feel informed, or would they feel deceived? If the answer is "they'd feel deceived" — even if you think they shouldn't — disclosure is the right call. Trust is built on the audience's expectations, not your rationalisations.
Exercise: Map Your AI Disclosure Zones
Give this a go: List 10 work tasks where you use (or could use) AI. For each, classify the AI involvement level (assistant/collaborator/creator) and the disclosure expectation in your context. Where do you currently disclose, and where don't you? Are there gaps between what you think is right and what you actually do?
Reflection: Most people find at least one area where their practice doesn't match their principles. That's normal — and identifying it is the first step toward aligning them.
Exercise: Disclosure Language Templates
Here's what we'd suggest: Draft 3-4 disclosure statements you could actually use in your work. Keep them natural, not legalistic. Examples to adapt:
- "I used AI tools to help research and structure this document, then edited and verified the content."
- "This analysis was developed with AI assistance for data processing; all conclusions and recommendations are my own."
- "AI helped generate initial drafts of sections X and Y; I wrote sections Z and W and edited everything for accuracy."
Reflection: Having pre-written disclosure language removes the friction of deciding in the moment. When disclosure is easy, people do it more.
Core Skill 2: AI and Job Impact
Let's address the elephant in the room: will AI take your job? The honest answer is more nuanced than either the utopian ("AI will just create better jobs!") or dystopian ("everyone will be unemployed!") narratives suggest. And understanding the realistic picture is genuinely important — not just for your career, but for how you lead teams, make hiring decisions, and prepare your organisation.
Last verified: March 2026
🔢A significant share of jobs — by some estimates over half — have substantial exposure to AI
meaning a meaningful portion of their tasks could be augmented or reshaped by AI tools, according to OECD analysis. But "exposure" doesn't mean replacement — fewer than 5% of occupations can be fully automated with current technology. The impact is task-level, not job-level.
Source: OECD Employment Outlook (ongoing series); McKinsey Global Institute future of work research
The augmentation vs. replacement framework
Last verified: March 2026
Rather than asking "will AI replace this job?", a more useful question is: "which tasks within this job will AI change, and how?" Most jobs contain a mix of:
- Automatable tasks — routine, data-heavy, pattern-matching. AI can do these faster and often better. Think: data entry, basic analysis, standard report generation, scheduling.
- Augmentable tasks — require human judgment but benefit from AI support. AI handles the heavy lifting while you direct and refine. Think: research synthesis, first-draft writing, code generation, customer analysis.
- Human-essential tasks — require empathy, nuanced judgment, relationship building, creative vision, or physical presence. AI can inform these but can't replace them. Think: strategic decisions, difficult conversations, creative direction, mentoring, negotiation.
⚖️ Job Impact Scenarios: Augmentation vs. Displacement
Scenario What Happens Who Benefits Who's at Risk Task automation Routine tasks eliminated; job restructured around higher-value work Workers who upskill to focus on judgment and creativity Workers whose role is primarily the automated tasks Productivity amplification Same work done by fewer people, or more work done by same team Top performers who leverage AI effectively Average performers in roles where output is easily measured New role creation Entirely new positions emerge (prompt engineering, AI oversight, ethics) Early adopters with cross-functional skills Those who wait for formal job descriptions before upskilling Quality elevation AI raises baseline quality; competitive advantage shifts to exceptional work Creative thinkers and domain experts Those who competed primarily on speed or volume of average-quality work
📈 Worker Attitudes Toward AI Job Impact (2025)
Source: PwC Global Workforce Hopes & Fears Survey (ongoing series); Gallup State of the Global Workplace research
Something worth considering: the anxiety about AI replacing jobs is not new — it's a pattern that has accompanied every major productivity technology. What is new is the speed. Previous waves of automation unfolded over decades, giving workers and institutions time to adapt. AI is compressing that timeline. The people who will navigate this best aren't necessarily the most technically skilled — they're the most adaptable and the quickest to integrate new tools into their existing expertise.
Exercise: Personal Task Audit
Give this a go: List 15–20 tasks you do regularly at work. For each, classify it as: automatable, augmentable, or human-essential. Be honest — resist the temptation to classify everything as human-essential.
What to observe: What percentage of your time goes to each category? If the automatable tasks disappeared tomorrow, what would you fill that time with?
Reflection: The purpose isn't to scare yourself — it's to see clearly where AI can free you up for higher-value work, and to start intentionally building skills in the human-essential category.
Exercise: Team Impact Assessment
Here's what we'd suggest: If you manage or work closely with a team, assess the AI impact across all team members' roles. Map which tasks across the team are automatable, augmentable, and human-essential. Then ask: if AI handled all the automatable tasks, how would the team's composition and focus need to change?
Reflection: Leaders who think about this proactively can guide their teams through transition rather than reacting when the changes are forced upon them.
Exercise: AI Concerns Conversation
Give this a go: Interview a colleague about their concerns around AI at work. The goal isn't to convince them of anything — it's to genuinely listen, then practise responding with empathy and honesty.
- Choose someone you trust — ideally someone whose feelings about AI differ from yours.
- Ask open-ended questions: "What worries you most about AI in our work?" and "What would make you feel more confident about how we're using it?"
- Listen without correcting. Take notes on what they say, especially the emotions behind the concerns.
- Respond honestly — acknowledge what you don't know, share your own uncertainties, and avoid dismissing their worries with "AI won't replace us" platitudes.
- After the conversation, write down three things you learned about how AI anxiety shows up in your workplace.
Reflection: We've found that the most common AI concern isn't "I'll lose my job" — it's "nobody is being honest with me about what's changing." Practising empathetic honesty is one of the most valuable skills you can build for navigating this transition alongside your team.
Core Skill 3: Responsible Deployment Decisions
Using AI personally is one thing. Deploying it in ways that affect other people — employees, customers, students, patients — raises the stakes considerably. If you're involved in decisions about how your organisation uses AI, you're making ethical choices whether you frame them that way or not.
Last verified: March 2026
🔢A large majority of consumers say they'd lose trust in a company
that used AI in ways they considered unfair or non-transparent — even if the AI-driven outcome was technically accurate. Perceived fairness matters as much as actual accuracy.
Source: Edelman Trust Barometer (ongoing series); Accenture responsible AI consumer research
The FATE framework for deployment decisions
Last verified: March 2026
When evaluating whether and how to deploy AI in a context that affects others, run through these four lenses:
- F — Fairness. Does this AI application treat all affected groups equitably? Have you tested for bias across demographic lines? Would you be comfortable if the affected group could see exactly how the AI made its decision?
- A — Accountability. Who is responsible when the AI makes a mistake? Is there a human in the loop for consequential decisions? Can the decision be appealed or overridden?
- T — Transparency. Do the people affected by this AI know it's being used? Can they understand — at least at a high level — how it works? Are they told when an AI-driven decision has been made about them?
- E — Ethics of impact. What are the second-order effects? Who benefits most? Who bears the most risk? Does this deployment widen or narrow existing inequities?
Knowledge Check
Your company is considering using AI to monitor employee productivity — tracking typing speed, application usage, meeting participation, and email response times. The AI would generate weekly "productivity scores" for each employee. What's your assessment?
Exercise: FATE Assessment of Your AI Tools
Give this a go: Choose an AI application your organisation uses (or is considering) that affects other people — customers, employees, students, patients. Run it through the full FATE framework. For each lens, rate the deployment as green (no concerns), amber (some concerns, mitigations needed), or red (serious concerns, pause and redesign).
Reflection: If any lens comes back red, that's a signal to slow down — not necessarily to stop, but to address the concern before scaling.
Exercise: AI in Hiring — Both Sides
Here's what we'd suggest: AI is now used on both sides of the hiring process. Explore both perspectives:
As a hiring team: What are the ethical implications of using AI to screen resumes, conduct initial interviews, or evaluate candidates? What biases might it introduce? What does fairness mean when an algorithm decides who gets an interview?
As a candidate: What are the ethical implications of using AI to write your cover letter, prepare for interviews, or negotiate your salary? Where's the line between "using a tool" and "misrepresenting your abilities"?
Reflection: Notice whether you hold candidates and companies to different standards. If AI-screened resumes feel unfair but AI-written cover letters feel fine (or vice versa), examine why.
Core Skill 4: Building an Ethical AI Culture
Individual ethics matter, but culture determines what actually happens at scale. A team where one person thinks carefully about AI ethics while everyone else doesn't is still an ethically fragile team. Building a culture where ethical AI use is the default — not the exception — requires intentional effort.
Last verified: March 2026
What ethical AI culture looks like in practice
Last verified: March 2026
- Psychological safety to raise concerns — people feel comfortable saying "I'm not sure we should use AI for this" without being dismissed as anti-technology
- Shared vocabulary for ethical discussions — the team has common frameworks (like FATE) so ethical discussions are structured, not just vibes
- Regular ethical check-ins — not one-off training, but ongoing conversations about how AI use is evolving and what new questions are emerging
- Leadership modelling — leaders who disclose their own AI use, acknowledge mistakes, and take ethical concerns seriously
- Clear, practical guidelines — not aspirational principles on a wall, but specific, actionable guidance for real scenarios
📈 What Drives Ethical AI Behaviour in Organisations
Source: Industry estimates based on Deloitte AI ethics research and MIT Sloan Management Review analysis
Accountability when AI makes mistakes
Last verified: March 2026
One of the thorniest ethical questions: when AI produces a wrong, biased, or harmful output and someone acts on it — who's responsible? The AI? The developer? The user? The organisation?
The emerging consensus: the human who deployed the output is accountable. AI is a tool; using a tool doesn't transfer responsibility to the tool maker. If you publish an AI-generated report with errors, the errors are your responsibility — just as a typo in a Word document is your responsibility, not Microsoft's. This principle has practical implications:
- Never blame AI for your outputs. "The AI wrote it wrong" is not a professional defence.
- Always review AI outputs before they reach their audience. Every output, every time.
- Build review processes proportionate to the stakes. Low-stakes internal emails need less review than client-facing reports or published content.
- Document your review process. If something goes wrong, you want evidence that you applied reasonable diligence.
Knowledge Check
An AI tool your team uses to generate customer emails sent an insensitive response to a grieving customer (the customer had mentioned a family bereavement, and the AI responded with a cheerful upselling message). The customer complains. Your manager asks what happened. How do you respond?
Here's a pattern we've noticed: the organisations with the strongest ethical AI cultures are the ones where mistakes are treated as learning opportunities, not career threats. When people are afraid to report AI failures, the failures don't disappear — they just go underground. A blame-free culture around AI mistakes isn't "going easy" on ethics; it's the only way to surface problems before they scale.
Exercise: Draft Your Team's AI Ethics Guidelines
Give this a go: Draft a one-page AI ethics guide for your team. Include:
- When AI use should be disclosed (use the three-tier framework from Core Skill 1)
- What data is never acceptable to share with AI tools (reference the classification framework)
- Who is accountable for AI-generated outputs
- How to report AI errors or concerns without blame
- How often the guidelines will be reviewed and updated
Reflection: The hardest part of this exercise is usually making the guidelines specific enough to be useful. "Use AI responsibly" means nothing. "Always review AI-generated customer emails before sending" means something.
Exercise: Ethical Grey Area Discussion Guide
Here's what we'd suggest: Prepare 5 ethical grey-area scenarios relevant to your work for a team discussion. Examples:
- Using AI to write a eulogy or sympathy message
- Using AI to prepare arguments for a negotiation where the other party doesn't have AI
- Having AI analyse a colleague's communication patterns to predict their behaviour
- Using AI to ghostwrite social media posts that present as authentic personal voice
- Using AI to generate interview questions that test for real skills vs. AI-assisted performance
Reflection: Grey areas don't have right answers — they have better and worse reasoning. The value is in the conversation, not the conclusion.
Exercise: FATE Assessment of a Specific AI Tool
Give this a go: Pick one AI tool your team actually uses day-to-day — a chatbot, a recommendation engine, an automated scheduling system, anything that touches real people. Then run a structured FATE assessment on it.
- Name the tool and describe what it does in one sentence.
- Fairness: Who does this tool affect? Are there groups who might be treated differently by it? If you can, test it with diverse inputs and note any patterns.
- Accountability: When this tool gets something wrong, what happens? Is there a clear person or process responsible for catching and correcting errors?
- Transparency: Do the people affected by this tool know it's being used? Could you explain — in plain language — how it makes its decisions?
- Ethics of impact: Who benefits most from this tool? Who bears the most risk? Has anyone asked the people most affected how they feel about it?
- Rate each lens as green, amber, or red. For any amber or red ratings, draft one concrete recommendation for improvement.
Reflection: We've found that most teams have never formally assessed the tools they already use — the focus tends to be on evaluating new deployments. But the tools already running quietly in the background are often where the biggest ethical blind spots live.
Challenge Exercises
These challenges put multiple ethical frameworks together in scenarios complex enough to resist simple answers. The goal isn't to find the "correct" response but to develop structured reasoning you can apply when real situations arise.
Challenge 1: AI Disclosure Policy for a Creative Agency
Scenario: You run a creative agency (copywriting, design, strategy). Your team increasingly uses AI in their workflows. Some clients explicitly ask whether AI is involved; most don't. Your competitors probably use AI but don't disclose it.
Task: Develop a comprehensive disclosure policy. Address: what level of AI involvement triggers disclosure, how to communicate it to clients, how to handle clients who object to AI use, how to price work that involves AI (should it cost less?), and how to maintain competitive position while being transparent.
Deliverable: A client-facing disclosure policy + internal implementation guidelines.
Success criteria: Is the policy honest without being competitively suicidal? Does it build trust rather than erode it? Would your best clients appreciate the transparency?
Challenge 2: AI Job Impact Communication Plan
Scenario: Your organisation is implementing AI tools that will automate significant portions of work currently done by a 30-person team. Some roles will change; a few may become redundant. Leadership asks you to develop the communication plan.
Task: Design a communication strategy that is honest about the impact without creating panic. Include: what to tell the team and when, how to frame the changes (augmentation vs. replacement — honestly), what support to offer affected employees (reskilling, redeployment, severance), how to maintain morale and productivity during the transition, and how to handle the inevitable rumour mill.
Deliverable: A phased communication plan with messaging for different audiences (leadership, affected team, broader organisation).
Success criteria: Would affected employees feel respected and informed? Does the plan balance honesty with compassion? Are the support offerings genuine, not performative?
Challenge 3: AI Ethics Committee Charter
Scenario: Your organisation wants to establish an AI Ethics Committee to oversee AI adoption and deployment decisions.
Task: Draft the committee charter. Define: purpose and scope, membership (who should be on it — and who shouldn't), decision-making authority (advisory vs. binding), review process for new AI deployments, escalation path for ethical concerns, reporting cadence and transparency requirements, and how the committee stays current as AI evolves.
Deliverable: A complete committee charter ready for leadership approval.
Success criteria: Does the charter have real teeth (not just advisory)? Is the membership diverse enough to represent different perspectives? Would employees trust this committee to protect their interests?
Challenge 4: Personal AI Ethics Statement
Scenario: You want to articulate your own principles for AI use — a personal code of ethics that guides your decisions when formal policies are absent or ambiguous.
Task: Write a personal AI ethics statement (500–800 words). Address: how you decide when to use AI, when and how you disclose AI involvement, what data you will and won't share with AI tools, how you ensure accountability for AI-assisted work, and how you balance efficiency with ethical considerations. Ground each principle in a real example from your experience.
Deliverable: Your personal AI ethics statement, written clearly enough that a colleague could understand and apply your principles.
Success criteria: Is it specific enough to guide real decisions? Does it acknowledge tensions and trade-offs honestly? Would you be proud to share it publicly?
Quick Reference
AI Attribution Framework
Last verified: March 2026
- AI as assistant (brainstorming, grammar, suggestions) — disclosure optional, mention if asked
- AI as collaborator (significant generated content, heavily edited) — disclosure recommended
- AI as creator (core output is AI-generated) — disclosure required
- Test: Would the audience feel deceived if they discovered AI involvement later?
FATE Framework for Deployment
Last verified: March 2026
- Fairness — does it treat all groups equitably? Tested for bias?
- Accountability — who's responsible when it goes wrong? Is there human oversight?
- Transparency — do affected people know AI is being used?
- Ethics of impact — who benefits, who bears risk, does it widen inequities?
Job Impact Assessment
Last verified: March 2026
- Automatable tasks: routine, data-heavy, pattern-matching — AI handles faster
- Augmentable tasks: judgment-required but AI-assisted — focus on directing and refining
- Human-essential tasks: empathy, creativity, relationships, complex judgment — invest here
- Action: Deliberately shift your time from automatable to human-essential work
Accountability Principles
Last verified: March 2026
- The human who deploys the output is accountable — AI is a tool, not an excuse
- Always review AI outputs before they reach their audience
- Review intensity should match the stakes (internal email vs. client report)
- Document your review process for high-stakes outputs
- Report AI errors without blame — learning requires honesty
Ethical AI Culture Checklist
Last verified: March 2026
- Does the team have shared frameworks for ethical discussion?
- Can team members raise ethical concerns without career risk?
- Do leaders model transparent AI use?
- Are AI guidelines specific and actionable (not just aspirational)?
- Is there a regular cadence for reviewing and updating AI practices?
- Are AI mistakes treated as learning opportunities?
Common Ethical Pitfalls
Last verified: March 2026
- "Everyone else is doing it" — other people's ethics are not your benchmark
- "The AI did it" — tools don't have agency; you chose to use the output
- "Nobody will know" — the question isn't whether you'll get caught; it's whether it's right
- "The efficiency justifies it" — efficiency is not an ethical argument when harm is involved
- "We'll fix it later" — ethical debt, like technical debt, compounds and is harder to fix at scale
When-to-Pause Checklist
- Am I about to deploy AI in a way that affects people who didn't consent?
- Would I be comfortable if my AI use practices were made public?
- Am I using AI for persuasion or manipulation without the other party's knowledge?
- Could this AI deployment widen existing inequities?
- Am I prioritising efficiency over fairness or transparency?
- Have I considered the perspective of the person most negatively affected by this decision?
The ethical skills you've developed in this path — navigating attribution decisions, understanding job impact honestly, making responsible deployment choices, and building cultures of ethical AI use — don't come with clean, permanent answers. The technology evolves, norms shift, and new grey areas emerge faster than policies can address them. What doesn't change is the underlying commitment: to use powerful tools with integrity, to be honest about trade-offs, and to consider the impact on people who don't have a seat at the decision-making table. We've found that the professionals who navigate AI ethics best aren't the ones with the most rules — they're the ones who've internalised the habit of asking "who does this affect, and would they consider it fair?" before they act. That habit, more than any framework, is what this path is really about.
Practice Project
Most teams are using AI without any shared understanding of what's okay and what isn't — and that ambiguity creates more anxiety than any policy ever would. This project gives you a way to start that conversation, even if you're not in a leadership role.
Time: 60 minutes
What you'll build: A Team AI Usage Policy — a practical, readable set of guidelines covering 5-7 policy areas, plus a decision flowchart for the grey areas that inevitably come up.
Why this matters: When we first tried to write AI guidelines for our own team, we discovered something interesting: the process of writing the policy was more valuable than the policy itself. The conversations it sparked — about attribution, about what data is too sensitive, about when AI-generated work needs disclosure — were conversations we should have been having months earlier. You might find the same thing.
Steps:
- Interview (or imagine) 3 team members about their AI concerns and wishes. If you can have actual conversations, spend 10 minutes with each person asking: "What worries you about how our team uses AI?" and "What do you wish we had clearer guidance on?" If interviews aren't practical, imagine 3 people with different roles and perspectives on your team — the enthusiast, the sceptic, and the pragmatist. Write down what each would likely say.
- Draft 5-7 policy areas. Start with the ones that came up in your interviews, then fill gaps. Common areas include: data handling and classification, attribution and disclosure, quality checking requirements, client-facing AI use, intellectual property considerations, training and upskilling expectations, and tools approval. You don't need to cover everything — focus on what actually causes confusion or conflict.
- Write clear, practical guidelines for each area. Avoid vague statements like "use AI responsibly." Instead, write guidelines people can actually follow: "AI-generated content for client deliverables must be reviewed by at least one team member before sending." Each guideline should answer the question: "If someone on my team reads this, will they know exactly what to do?"
- Add a decision flowchart for edge cases. Create a simple yes/no flowchart for the situations your guidelines don't fully cover. Start with: "Does this involve sensitive or confidential data?" and branch from there. 5-7 decision points is usually enough to handle most grey areas without overcomplicating things.
Deliverable: A draft AI usage policy — clear enough to share with your team for feedback, practical enough to actually follow.
Stretch goal: Present your draft to your team or manager. Frame it as a conversation starter, not a finished document: "I drafted this based on some common questions — what would you add or change?" The feedback you get will make the final version significantly better.
Reflection: Which policy area generated the most uncertainty for you? That's probably the area where your team most needs shared norms — and where having even an imperfect guideline would reduce daily friction.
A policy document gathering dust in a shared drive helps nobody. What matters is the shared understanding it creates — the moment when everyone on your team knows the ground rules and can make faster, more confident decisions about how they use AI. You don't need permission to start that conversation. You just need a draft worth discussing.