Leadership & Communication · Leading in the AI era: agents, async and responsible AI · lesson 21 of 21 · 15 min

Using AI in everyday management, responsibly

A capable assistant, not a substitute for judgement

AI assistants can save managers hours each week: drafting messages, summarising documents, preparing for one-to-ones, turning notes into action lists, analysing anonymous survey comments and rehearsing difficult conversations. The risk is not that AI is useless; it is that it is fluent. A polished paragraph can hide a wrong fact, a biased assumption or a tone that is not yours, and your team will judge you, not the tool.

This lesson gives you a practical rule set: where AI helps, where to be careful, and where it should not be used without strong safeguards.

A traffic-light guide for managers

| Green: use freely (with checks) | Amber: use with care and safeguards | Red: do not use AI to decide | |---|---|---| | Drafting routine messages and agendas | Summarising meetings (announce it; human checks) | Hiring, promotion, pay, termination decisions | | Summarising long documents and threads | Analysing anonymous engagement survey comments | Performance ratings or disciplinary outcomes | | Preparing questions for a 1:1 or coaching | Drafting feedback from your own notes | Monitoring individuals' productivity or emotions | | Rehearsing a difficult conversation | Writing job descriptions (check for biased language) | Anything involving health, grievances or protected characteristics | | Structuring a presentation or plan | Translating messages (check with a fluent speaker) | Pasting confidential data into unapproved tools |

"Red" does not mean AI can never be involved in those processes; it means a person makes the decision, with proper process, and any AI tool used has been approved, assessed and disclosed. Many organisations have HR, legal and data protection processes for exactly this.

Why the red zone matters

Fairness and bias. AI systems can reproduce biases present in their training data or in the way a prompt is framed. A ranking of candidates or a summary of "performance issues" can quietly disadvantage groups.
Law. Data protection law restricts certain solely automated decisions with legal or similarly significant effects on people (for example Article 22 of the UK GDPR and EU GDPR). The EU AI Act lists AI used for recruitment, promotion, termination, task allocation and monitoring or evaluating workers as high-risk, with obligations that are being phased in (their application date was postponed by 2026 amendments; check the current timetable). Other countries have their own rules; check locally.
Trust. People accept tough decisions more readily when they believe a fair human process was followed. Discovering that a machine wrote your appraisal destroys that belief.

Transparency with your team

Tell your team how you use AI in management: for example, "I use our approved assistant to draft agendas and summarise documents; I don't use it to evaluate your performance." Agree team norms (see the team working agreement) and follow your organisation's AI policy. When AI has materially shaped something people rely on, such as a summary of their survey comments, say so.

Protecting data

Use only tools approved by your organisation for the data involved. Do not paste personal data about team members (health, family circumstances, performance concerns, grievances) into consumer tools. Anonymise where you can: roles, not names. Check whether your tool's settings allow inputs to be used for training, and follow your organisation's guidance.

Worked example

Illustrative. James manages a ten-person finance team in Manchester. Before year-end reviews he uses the company's approved assistant to turn his own dated notes (kept throughout the year) into a first-draft structure for each review: achievements, development areas, examples. He then checks every example against his notes, removes a sentence the tool invented about "communication issues", adds context only he knows, and rewrites the opening in his own voice. The rating is decided by him and calibrated with other managers, with no AI involvement. Preparation time falls, and the reviews are more specific than last year's.

Hands-on: a personal AI use policy and a safe drafting workflow

1. Write your personal AI use statement (share it with your team):

How I use AI as a manager
- I use [approved tool] to: draft agendas and routine messages; summarise long documents;
  prepare coaching questions; rehearse difficult conversations.
- I check every fact and rewrite in my own voice before sending.
- I do NOT use AI to: rate performance, decide pay, promotions or discipline,
  or monitor individuals.
- I never paste personal, health or grievance information into AI tools.
- If you're unsure how AI was used in something that affects you, ask me.

2. Safe drafting workflow for feedback or review text:

Step 1  Write your own bullet notes first (dated, specific, SBI format).
Step 2  Anonymise: roles not names; remove health/family/grievance details.
Step 3  Prompt (approved tool only):
        "Turn these notes into a clear, balanced draft in plain British English.
         Use only the facts given. Do not add examples, judgements or adjectives
         that aren't in the notes. Flag anything that sounds vague or harsh."
Step 4  Check every sentence against your notes; delete anything added.
Step 5  Rewrite the opening and closing in your own words.
Step 6  Deliver in a conversation, not as a pasted document.

3. Survey comment analysis (amber): remove names and identifying details first; ask for themes with counts and representative (paraphrased) quotes; then read a random sample of raw comments yourself to check the themes are real.

Measuring success

Time saved on preparation; fewer revisions to your messages; team trust (ask in a pulse survey whether people feel AI is used fairly); and zero data incidents.

Common mistakes

Sending AI drafts without checking facts and tone.
Letting AI "fill gaps" in performance evidence.
Using consumer tools for personal or confidential information.
Hidden use that people later discover.
Treating an AI ranking or summary as objective.

Quick self-check

Look at the last five things you used AI for at work. Which colour zone was each in? Did you apply the right safeguards?

Video lecture: Using AI in everyday management, responsibly

Lecture coming soon · 13 chapters · about 9 minutes. Read the full transcript below.

Using AI in everyday management
Why it matters
Traffic-light guide
Satnav and licence
Why the red zone matters
Transparency and data
Worked example 1: Asad in Islamabad (illustrative)
Worked example 2: James in Manchester (illustrative)
Safe drafting workflow
Watch me: AI use statement + audit
Measuring success
Common mistakes
Recap and try this now

Lecture transcript

Using AI in everyday management

Here's a scene that's playing out in offices everywhere. A manager, short of time, asks an AI assistant to write a team member's annual review. It produces three beautifully written paragraphs in seconds. One of them mentions communication issues that never happened. The manager doesn't notice. The team member does. And from that moment, nothing that manager says about performance is fully trusted. AI can save managers hours every week. But its fluency is exactly what makes it risky. In this lecture you'll learn a traffic-light guide for AI in management, why the red zone matters, how to be transparent, how to protect data, and a safe drafting workflow.

Why it matters

Why does this matter? Because managers are among the heaviest potential beneficiaries of AI, and among the most exposed to its risks. So much of management is writing, summarising and preparing: agendas, updates, one-to-one notes, feedback, plans. AI genuinely helps with all of that. But management also involves decisions about people's careers, pay and wellbeing. Get those wrong, or let people believe a machine made them, and you damage fairness, you may break the law, and you lose trust that takes years to rebuild.

Traffic-light guide

Here's the concept: a traffic-light guide. Green: use freely, with checks. Drafting routine messages and agendas, summarising long documents, preparing coaching questions, rehearsing a difficult conversation, structuring a plan. Amber: use with care and safeguards. Summarising meetings, analysing anonymous survey comments, drafting feedback from your own notes, writing job descriptions, translation. Red: don't use AI to decide. Hiring, promotion, pay, termination, performance ratings, discipline, monitoring individuals, and anything involving health or grievances. Here's the key idea: red doesn't mean AI is banned from the building. It means a person makes the decision, through a proper process, with any tool approved and disclosed.

Satnav and licence

Here's an analogy. Think of a satnav and a driving licence. The satnav is brilliant. It plans routes, spots traffic and saves you time. But if it tells you to turn into a pedestrian street, you don't do it, and if you do, it's your licence, not the satnav's, on the line. AI is your management satnav. Follow it gladly on the motorway of routine drafting. Question it in unfamiliar streets. And never let it drive through the red lights of people decisions. The responsibility always stays with the person holding the licence.

Why the red zone matters

Why is the red zone so important? Three reasons. Fairness: AI systems can reproduce biases from their training data or from how a prompt is framed, and a ranking or summary can quietly disadvantage a group. Law: data protection law, such as Article twenty-two of the UK and EU GDPR, restricts certain solely automated decisions with significant effects on people. And the EU AI Act lists AI used for recruitment, promotion, termination, task allocation and worker monitoring as high-risk, with obligations being phased in. Check the current timetable and your local law. And trust: people accept tough decisions more readily when they believe a fair human process was followed.

Transparency and data

Next, transparency and data. Tell your team how you use AI. A short statement works: I use our approved assistant to draft agendas and summarise documents. I don't use it to evaluate your performance. If you're unsure how AI was used in something that affects you, ask me. That one paragraph prevents the damaging moment when someone discovers hidden use. And protect data. Use only tools your organisation has approved for the data involved. Never paste health information, family circumstances, grievances or performance concerns into consumer tools. Anonymise: roles, not names.

Worked example 1: Asad in Islamabad (illustrative)

A simple worked example. Asad runs a small accounting firm in Islamabad with seven staff. He uses an approved assistant each Monday to turn his rough notes into a clear weekly plan and an agenda for the team meeting: green zone. When he wants to understand a staff survey, amber zone, he removes names and identifying details first, asks for themes with counts, then reads a random sample of raw comments himself. One theme the tool called workload stress turns out, in the actual comments, to be mostly about unclear priorities. He fixes priorities, not hours.

Worked example 2: James in Manchester (illustrative)

Now a realistic business scenario, illustrative. James manages a ten-person finance team in Manchester. Before year-end reviews, he uses the company's approved assistant to turn his own dated notes, kept throughout the year, into a first-draft structure for each review. He instructs it to use only the facts given. Then he checks every sentence against his notes. In one draft, he finds a sentence about communication issues that isn't in his notes at all. He deletes it. He rewrites openings in his own voice. The ratings are his, calibrated with other managers, with no AI involvement. Preparation time falls, and the reviews are more specific than last year's.

Safe drafting workflow

Let's make the safe drafting workflow explicit, because it applies to feedback, reviews and any sensitive message. Step one, write your own notes first: dated, specific, in situation, behaviour, impact form. Step two, anonymise. Step three, use an approved tool and tell it to use only the facts given, not to add examples, judgements or adjectives, and to flag anything vague or harsh. Step four, check every sentence against your notes and delete additions. Step five, rewrite the opening and closing in your own words. Step six, deliver it in a conversation, not as a pasted document.

Watch me: AI use statement + audit

Watch me do it. I write my personal AI use statement for an illustrative team. I use our approved assistant to draft agendas and routine messages, summarise long documents, prepare coaching questions and rehearse difficult conversations. I check every fact and rewrite in my own voice. I don't use AI to rate performance, decide pay, promotions or discipline, or to monitor individuals. I never paste personal, health or grievance information into AI tools. If you're unsure, ask me. Then I classify my last five uses. Four are green. One, summarising a team retrospective, is amber, and I realise I didn't tell the team. I'll mention it at the next meeting.

Measuring success

How do you know you're using AI well? Measure time saved on preparation. Look at how often you have to correct or retract something you sent. Ask in a pulse survey whether people feel AI is used fairly and transparently in the team. And aim for zero data incidents: nothing personal or confidential pasted into tools that shouldn't have it. If time saved is high but trust is falling, slow down. Trust is the harder thing to rebuild.

Common mistakes

Common mistakes. Sending AI drafts without checking facts and tone. Letting AI fill gaps in performance evidence with plausible sentences. Using consumer tools for personal or confidential information. Using AI secretly, and being found out. Treating an AI ranking or summary as objective, when it reflects its data and your prompt. Using AI to monitor individuals' activity or mood. And outsourcing empathy: an AI-written apology or thank-you that you don't personally mean.

Recap and try this now

Let's recap. AI can save managers hours, but its fluency hides errors, bias and the wrong tone. Use the traffic-light guide: green with checks, amber with safeguards, and red for people decisions that humans must make through proper process. Be transparent, protect data, and use the six-step safe drafting workflow. Your try this now: write your personal AI use statement using the template in the lesson, share it with your team, and classify your last five uses of AI by colour. That completes this course. Well done. Now pick one template from each module and use it this month.

Key takeaways

AI saves managers time on drafting, summarising and preparation, but fluency can hide errors, bias and the wrong tone.
Use a traffic-light rule: green for routine drafting, amber with safeguards, red where people decisions must be made by humans.
Employment decisions carry legal, fairness and trust risks; data protection and AI laws apply.
Be transparent with your team about how you use AI, and never paste personal or grievance data into unapproved tools.
Write your own notes first, constrain the tool to your facts, check every sentence, and deliver feedback in person.

Try it

Write and share your personal 'How I use AI as a manager' statement, then classify the last five things you used AI for as green, amber or red.