Keep an AI-assisted performance-review prompt to the minimum evidence needed for the job-related question. Leave out unnecessary medical or personal details, private information about other people, guesses about protected traits, unchecked monitoring data, confidential material in an unapproved service, and instructions that ask the model to decide pay, promotion, discipline, or termination. Use an approved tool, verify every factual statement it produces, and keep the final assessment with an accountable human.
The rule that governs every review prompt
Before pasting anything into an AI system, define the review purpose in one sentence: for example, “Summarise documented project-delivery evidence for the January–June review period against the published expectations for this role.” Then include only information that is necessary for that purpose, directly related to the work, current enough to be reliable, and authorised for the manager to use.
As an Amazon Associate I earn from qualifying purchases.
This is practical cross-jurisdiction guidance, not a universal legal rule. What an employer may collect, how it may process employee information, and which AI uses are regulated depend on the country or state, the purpose, the tool’s configuration, collective agreements, and internal policy. The UK Information Commissioner’s Office (ICO) worker-monitoring guidance is currently under review following the Data (Use and Access) Act, so UK teams should check the current page and their organisation’s advice before operationalising a workflow.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →A useful test is the “remove it and see” test: if deleting a data item would not prevent the model from organising verified, job-related evidence, delete it. Extra context can increase privacy, discrimination, and accuracy risks without improving the review.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
1. Sensitive personal and medical details that the review does not need
Do not paste diagnoses, treatment histories, medication, therapy notes, disability details, fertility or pregnancy information, family circumstances, bereavement, financial hardship, or other sensitive personal information simply to make a prompt appear complete. A person’s private situation may be relevant to a properly handled accommodation or leave process, but it does not automatically belong in a performance-review prompt.
The Australian Office of the Australian Information Commissioner (OAIC) recommends, as a best practice, not entering personal information—particularly sensitive information—into publicly available generative-AI products. The ICO’s data-minimisation principle similarly says worker-monitoring information should be limited to what is necessary for the stated purpose.
Separate performance evidence from support processes
If an employee has requested an accommodation, route that request through the designated HR, occupational-health, or accessibility process. In the review prompt, describe the agreed work expectations and any approved adjustment only to the extent needed to evaluate performance fairly; do not include the underlying diagnosis or medical narrative.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
For example, a manager could write that a revised deadline was approved as part of a documented workplace adjustment. The manager should not add the condition, treatment, doctor’s name, or private correspondence. The review should assess agreed outcomes under the adjusted arrangement, not invite an AI system to judge whether a health condition explains a result.
2. Private information about coworkers, customers, or subordinates
Names and identifying details of other people are usually unnecessary for an individual review. Remove customer names, personal contact details, coworker complaints, subordinate records, private messages, and details that would let the model identify a person. Replace them with role-based descriptions or de-identified references when the relationship is genuinely needed to understand the evidence.
De-identification is not a magic guarantee. A rare incident, a distinctive job title, a date, or a small team can make someone identifiable when combined. Use the least detail that preserves the work fact, and confirm that your organisation permits the processing. If a complaint or customer case is essential evidence, use the approved case-management or HR process rather than copying an entire private record into a general-purpose service.
What can remain
Keep a neutral description of the work event, its date or review-period position where necessary, the expected standard, and the observable result. “A support escalation remained unresolved for 48 hours after the documented service-level target” is more useful than a customer’s name, phone number, account history, or emotional language.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minute3. Requests to infer protected or sensitive traits
Never ask an AI model to infer disability, health, race, ethnicity, religion, sex, gender identity, sexual orientation, age, pregnancy, or another protected or sensitive characteristic from work patterns, writing style, absences, location, keystrokes, communication habits, or other proxies.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
The ICO explains that a model can reproduce discrimination through proxy patterns even when the protected field is absent. An inference can also become sensitive personal data when it reveals or predicts a protected characteristic. Removing a column called “disability” does not make a dataset fair if the model can reconstruct a similar signal from leave patterns, schedule changes, language, or task history.
Prompts that cross the line
- “Does this employee appear to have a health condition based on absences and slower response times?”
- “Infer whether the writer is likely to be from a particular ethnic or religious background.”
- “Identify people whose communication suggests a disability or mental-health issue.”
- “Use attendance patterns to estimate who may become pregnant or need leave.”
These requests are not made safe by adding “for fairness,” “for risk management,” or “do not be biased.” The underlying task is still an unsupported sensitive inference.
Ask about observable work instead
Replace trait speculation with evidence questions: “Which documented deliverables met the role’s quality standard?” “Where is evidence missing?” “What clarification would the employee need to respond to this feedback?” The model may organise facts; it should not diagnose a cause or label a person.
4. Unverified monitoring data, bare scores, and misleading proxies
Do not treat keystroke counts, mouse movement, login duration, webcam flags, badge swipes, “active” time, sentiment scores, availability indicators, or opaque productivity ratings as self-proving facts. Monitoring systems measure particular signals and miss other work, such as planning, mentoring, deep concentration, offline work, accessibility-related variations, or tasks performed in another approved system.
The ICO warns that monitoring-derived information must not be incorrect or misleading and that analytics can make incorrect inferences. Employees should be able to see and challenge information used in performance reviews under the ICO guidance. A prompt that turns an unexamined dashboard score into a statement about effort or commitment can launder an unreliable inference into official feedback.
Validate before any AI use
- Identify the measurement. Record what the system actually counts, its time zone, sampling rules, exclusions, and known blind spots.
- Check the source. Confirm that the data came from the correct person, period, team, and system, and that duplicate events or outages were handled.
- Test the interpretation. Ask whether the metric measures the performance outcome or merely a proxy for activity.
- Give notice and access where required. Follow the applicable monitoring policy and provide the route for an employee to inspect or challenge information.
- Use corroboration. Pair a metric with documented deliverables, quality checks, customer outcomes, or other job-related evidence.
If the underlying record cannot be explained or challenged, do not ask an AI model to convert it into a rating.
5. Instructions that make the model decide an employment outcome
A manager should not prompt an AI system to “decide who should be fired,” “rank employees for promotion,” “set a pay score,” “select the lowest performer,” or otherwise determine a consequential employment action from a prompt. Employment evaluation and decisions affecting work relationships are identified as potentially high-risk AI uses in European Commission AI Act Service Desk guidance. U.S. Equal Employment Opportunity Commission (EEOC) and Department of Justice materials also describe disability-discrimination risks when software is used for monitoring, pay, promotion, or other employment decisions.
Free tools Windows power users keep installed
One-click scans. No signup required.
The risk is not removed by saying that a human will “rubber-stamp” the answer. A consequential recommendation can anchor the manager, hide missing evidence, and make it difficult for the employee to understand or challenge the basis for the action.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Keep the decision outside the model
Use AI, if approved, for bounded support tasks such as organising evidence by competency, finding contradictions, listing missing documentation, or drafting neutral questions for a conversation. The manager and authorised decision-makers must evaluate the complete record, apply policy, consider context, document reasons, and make the final decision. Do not upload a roster and ask for a ranking when the intended result is a pay, promotion, discipline, or termination decision.
6. Confidential employer or third-party information in an unapproved service
Never paste internal strategy, unreleased financial information, source code, security details, credentials, customer records, employee files, contract terms, investigation material, or other confidential content into a public or unapproved AI service. “The tool is free” or “the prompt is deleted from my screen” does not establish that the data is private.
Check the employer’s approved-tool policy and the actual vendor configuration. Important questions include whether prompts and outputs are retained, whether they are used to improve a model, who can access them, where processing occurs, how long logs remain, whether a subcontractor receives them, and how deletion requests work. Contractual commitments and product settings can differ from a consumer-facing default.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Minimise before submitting
- Use the organisation’s approved enterprise workspace, not a personal account.
- Remove identifiers and confidential fields that are not needed for the task.
- Confirm the permitted data classification and retention setting.
- Keep the original evidence in the authorised system of record.
- Do not upload credentials, access tokens, secrets, or complete employee files.
The OAIC notes that privacy obligations apply to personal information entered into AI products and recommends against entering personal or sensitive information into publicly available generative AI. If no approved configuration exists for the data, do not submit it.
A safer prompt pattern
A practical prompt uses a de-identified employee reference where workable, a defined review period, published role expectations, and a small set of current, job-related examples that the manager is authorised to use. It asks the model to organise evidence and expose uncertainty—not to guess causes, traits, or outcomes.
Purpose: Organise evidence for a mid-year review against the documented expectations for the role of [role title].Review period: [start date] to [end date].Evidence: The following verified, job-related examples are authorised for this review: [de-identified examples with dates, expected standard, and observed result].Tasks:1. Group each example under the relevant documented expectation.2. Distinguish observed facts from interpretations.3. Flag missing, contradictory, or ambiguous information.4. Draft neutral questions the manager should resolve before writing feedback.5. Draft concise, respectful language that does not infer medical, demographic, or other protected traits.Do not make a rating, ranking, pay recommendation, promotion recommendation, disciplinary recommendation, or termination recommendation. Do not infer causes or personal characteristics. Quote or paraphrase only the evidence supplied.
The bracketed fields in this example are template variables, not an invitation to add every available record. Populate them only with information that passes the necessity, authorisation, accuracy, and approved-tool checks.
How to run the workflow without surrendering accountability
Step 1: Define the decision owner and purpose
Write down what the review is for, who is accountable for it, and which policy or competency framework applies. If the real objective is to select someone for dismissal or compensation, do not disguise that objective as a neutral “summary” task; route it through the required human and HR process.
Recommended Free Tools
Step 2: Build a minimum evidence set
Collect the role expectations and a limited set of dated, verifiable examples. Prefer outcomes, quality checks, documented commitments, and prior feedback over personality descriptions. Exclude material that is merely interesting, embarrassing, or predictive of a protected trait.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
Step 3: Check people and confidentiality
De-identify the employee and third parties where practical. Remove customer and coworker details, medical information, credentials, confidential strategy, and data outside the manager’s authorisation. Confirm that the selected AI service is approved for the classification of information being processed.
Step 4: Ask for organisation, not judgment
Request a fact-to-expectation mapping, a list of gaps, and neutral questions. Instruct the model to mark uncertainty and to avoid inferring a cause. Keep any proposed wording separate from the official review record until a human has checked it.
Step 5: Verify every output against the source
Compare each factual statement with the original record. Check names, dates, quantities, status, quotations, and whether the model turned a possibility into a fact. Delete invented examples, unsupported explanations, and conclusions based only on a proxy metric.
Step 6: Add context the tool cannot know
Consider approved adjustments, changes in scope, system outages, staffing constraints, dependencies, prior notice, and the employee’s response. Include only context that is authorised and relevant; do not add private medical narratives to explain it.
Step 7: Give the employee a meaningful chance to respond
Follow the applicable policy for sharing evidence and allowing correction. The ICO guidance emphasises transparency, accuracy, and challengeability for information used in worker monitoring and reviews. A person should not be judged on an opaque AI summary they cannot inspect or contest.
Step 8: Make and document the human assessment
The manager or authorised panel decides what the evidence means under the organisation’s framework. Record the evidence relied on, material context considered, corrections made to the AI draft, and the human reason for the final assessment. Preserve the source records according to the approved retention rules.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Bad prompt versus safer rewrite
| Risky instruction | Why it is unsafe | Safer direction |
|---|---|---|
| “Read these attendance logs and identify who is probably dealing with a health issue.” | Requests a sensitive health inference from a proxy. | “Summarise verified attendance records only where they affect a documented coverage commitment, and flag gaps without inferring a cause.” |
| “Rank the team from best to worst for promotion.” | Delegates a consequential employment decision and may hide unequal evidence. | “Map each person’s documented achievements to the published promotion criteria and list evidence that is missing for human review.” |
| “Use customer emails, names, and complaints to write a blunt review.” | Exposes third-party personal information and invites unsupported character judgments. | “Summarise de-identified, authorised service outcomes against the role standard using neutral language.” |
| “This dashboard says 62% activity. Explain why the employee is disengaged.” | Treats an unexplained proxy as fact and asks for a causal inference. | “Describe what the metric measures, identify limitations, and compare it with verified deliverables before drafting questions.” |
| “Paste the confidential investigation file and decide whether termination is justified.” | Combines unapproved disclosure with a request for a high-impact decision. | “Use the authorised HR process and ask a human decision-maker to apply the relevant policy to verified findings.” |
Checks before pressing Submit
- Purpose: Is the prompt tied to a stated, job-related review question?
- Necessity: Would the task still work if each personal detail were removed?
- Authority: Is the manager authorised to use every item and to send it to this service?
- Accuracy: Are dates, ownership, definitions, and measurement limits verified?
- Third parties: Are coworker, customer, and subordinate identifiers removed unless strictly necessary and authorised?
- Sensitive traits: Does the prompt avoid inferring health, disability, race, religion, sex, age, or other protected characteristics?
- Consequences: Does it prohibit ratings, rankings, pay, promotion, discipline, and termination decisions by the model?
- Transparency: Can the employee understand and challenge the information used, where applicable?
- Tool settings: Are retention, access, reuse, location, and deletion terms acceptable under policy?
- Human review: Is a named person responsible for checking the output and making the assessment?
What to do when an unsafe prompt has already been used
Stop further processing and do not copy the output into the personnel record. Preserve enough information for the organisation’s privacy or security team to investigate, including which service was used, what categories of data were submitted, when, and who could access the account. Follow the employer’s incident-reporting procedure; do not attempt to conceal the event by deleting evidence if policy requires preservation.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Ask the tool administrator or privacy lead whether retention or deletion controls can be applied, and determine whether any third-party or employee notification is required. Rebuild the review from authorised, minimised evidence in the approved workflow. Treat the generated text as untrusted until every factual claim has been checked.
Questions managers commonly ask
Can I include an employee’s disability if it explains missed targets?
Do not include the diagnosis or medical details merely to explain a result. Use the approved accommodation or leave process, and evaluate documented expectations under any agreed adjustment. The U.S. EEOC’s small-business ADA primer emphasises that actual job performance is a better measure of ability than disability and describes limits on disability-related inquiries.
Is a de-identified dataset automatically safe?
No. Rare events and combinations of role, dates, and circumstances can re-identify people, and a model can still infer protected traits from proxies. Minimise the data, use the approved service, and test whether the remaining details are necessary.
Can AI draft the wording of feedback?
It can, when the tool and data are approved and the task is limited to neutral drafting from verified evidence. A manager must check accuracy, add relevant context, remove unsupported claims, and own the final wording. Do not let a polished draft substitute for a fair assessment or a conversation.
What if the model produces a confident explanation?
Confidence is not evidence. Trace the statement to the original record. If the source does not establish it, label the information missing or remove the claim. Never use an AI-generated cause, trait, or prediction as a fact in a review.
Should monitoring data be excluded entirely?
Not necessarily. A monitoring measure may be relevant if its purpose, accuracy, limits, notice, retention, and challenge process are established and the measure is genuinely connected to the job. Validate it and corroborate it with outcome evidence; do not treat a bare score as a complete performance assessment.
Does a human sign-off make any AI use acceptable?
No. Human involvement does not cure unnecessary disclosure, unlawful inference, inaccurate data, or an unapproved service. Human accountability is the final safeguard after minimisation, accuracy, fairness, transparency, and tool-governance checks—not a replacement for them.
Bottom line for managers
An AI review prompt should contain a narrow purpose, verified job evidence, the relevant expectations, and nothing more. Keep out sensitive personal details, other people’s private information, protected-trait guesses, unexplained monitoring scores, confidential data in unapproved tools, and instructions that hand an employment decision to a model. Let AI organise and question the record if policy allows; let an accountable human verify the facts, hear the employee’s response, apply the organisation’s rules, and make the assessment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
