Units
1. Writing a Job Announcement and Position Description 2. CV Screening, Summarization and Risk of Bias 3. Interview Questions and Structured Assessment 4. Candidate Experience and Recruitment Communication 5. Orientation and Onboarding Content 6. Training and Development (L&D) Material Production 7. Performance Feedback and Evaluation Drafts 8. HR Policy and Internal Communication Texts 9. Employee Surveys and Free Text Analysis 10. KVKK, Data Privacy and Safe AI Use 11. AI Governance in HR: Ethics, Boundaries and Validation
Unit 7 / 11

Performance Feedback and Evaluation Drafts

Gains:

  • Ability to produce constructive performance feedback drafts based on concrete observations
  • Ability to structure balanced, fair and development-oriented feedback with SBI, STAR and similar frameworks
  • Draft of AI in feedback, being able to retain the role of human decision and responsibility

Performance feedback is the most difficult task for managers. Most people can give positive feedback; The challenge is to communicate areas of improvement in a constructive, concrete and fair manner. Poorly worded feedback can demotivate, create defensiveness, or even lose an employee. Artificial intelligence (AI) is a powerful aid in framing feedback correctly and softening language. But the unchanging principle of this unit is this: the content of the feedback should be based on real observation, and the decision and responsibility should remain with the person. AI writes draft; A human gives and owns the feedback.

A few terms: Feedback is communication that reflects the impact of a behavior to a person. A performance review is a formal review of a period; It usually results in promotions and bonuses. SBI is a framework that structures feedback on the Situation-Behavior-Impact axis. STAR is a structure that describes an event as Situation-Task-Action-Result and is used in both interviews and evaluation.

Constructive Feedback with SBI Framework

Effective feedback targets the behavior, not the person. “You are irresponsible” is a judgment and triggers defence; "The fact that the last three reports were left until the last minute forced the team" is an observation and can be discussed. Give the AI ​​your observations and frame it. Step by step process:

  1. Collect observations. Concrete events, dates, effects. (This step is entirely your responsibility.)
  2. Select the frame. SBI, STAR or “strengths + room for improvement” balance.
  3. Have the draft produced. In a behavior-focused, non-blaming language.
  4. Fairness and fact checking. Is the draft factual and balanced?
  5. Prepare for the interview. Possible reactions and answers.
  6. Give the feedback. Man directs the conversation; AI does not engage in conversation.

Your role: an experienced HR business partner.Task: Transform the following observations into a constructive feedback outline within the SBI framework (Situation-Behavior-Impact).Tone: respectful, growth-oriented, non-accusatory.Keep each point concrete; focus on behavior, not personality. Give both strengths and development areas in balance.<observations>- The last 3 project reports were left to the last day, the team had to wait- The client was very prepared in his presentation, positive feedback came (satisfaction 9/10)- He interrupted his colleague's opinion twice in the team meeting</observations>

Tip: Include both strengths and areas for improvement in the feedback. Feedback that only lists the negative feels unfair; If someone only speaks positive things, it will not lead to development. In the example above, including positive presentation performance establishes this balance. But don't mechanicalize the "praise sandwich" (a single criticism sandwiched between two praises); employees notice this pattern.

The Most Critical Rule: Observation Comes From You

AI cannot produce performance data; It simply expresses the observation you entered. Asking for feedback without observing leads the model to make up claims that seem reasonable but are not true — this is both unfair and dangerous. Shaping an employee's career just because "AI wrote this" is unacceptable.

Correct flow:1) YOU: enter concrete observations (date, event, impact, figures if any)2) AI: translate them into constructive language and framework3) YOU: review the draft, confirm its fidelity4) YOU: give feedback and lead the conversation

step

Owner

What it is (observation/data)

Manager / HR

How to express (draft)

A.I.

Confirmation of truth and justice

Manager / HR

feedback conversation

Manager (human)

Results (promotion, bonus, PIP)

Manager (human)

Preparing for Difficult Conversations and Self-Assessment

AI also helps you prepare for the feedback conversation. Rehearsing a difficult conversation in advance will help you stay calm in the real conversation:

I think the employee may be defensive when giving this feedback. Write down 3 possible reactions and suggest a constructive, calm response to each. My goal is not to blame, but to find a solution together. Also, suggest a neutral opening sentence that will open the conversation.

You can also help the employee structure his or her self-assessment:

Your role: a career coach. Task: Prepare a guiding template with 6 questions that will help an employee write an end-of-term self-evaluation. The questions should use the STAR structure (Situation-Task-Action-Result) and question achievements and development areas in a balanced way.

Three Mini Cases

Case 1 — From harshness to constructiveness. "You're always late, you're unreliable," one manager's sketch said. AI translated the same observation with SBI into "the deadline for the last 3 deliveries disrupted the team's plan; let's establish a delivery rhythm together." Instead of getting defensive, the employee talked about a solution; In the next quarter, the number of late deliveries decreased from 3 to 0.

Case 2 — Consistent calibration. In a team of 12 people, each manager wrote differently; Some called everyone "perfect", some did not like anyone. HR had AI prepare a common evaluation rubric and sample statement bank. In the calibration meeting, the discrepancy between scores decreased significantly and the number of objections halved.

Case 3 — Saving time. Term evaluation of a department of 30 people took the manager 2 full days. AI has halved the time by translating each employee's actual observations into the SBI draft; The manager devoted the time he saved to personalizing drafts and conducting interviews. Quality has not decreased, on the contrary consistency has increased.

Weak Prompt / Strong Prompt

Weak prompt:Write a performance evaluation for this employee.

Result: Because AI doesn't know actual performance, it produces generic, hollow, or even made-up statements; unfair and risky.

Strong prompt: [actual observations (date/event/impact/number) + SBI or STAR framework + tone + “focus on behavior” + “balance strengths and area of development” + “don't add anything you're not sure about” constraint]

The result: a constructive, balanced, directly usable outline based on your observations.

Attention: Do not forward the AI-generated draft to the employee as it is, without reading it at all. Even if the outline is based on your observation, whether the language is appropriate, fair, and personalized to the employee's context and your relationship—only you know that. The final performance evaluation and its consequences (promotion, bonus, performance development plan) are always the responsibility of the manager. This is the same principle of safety-critical approval in engineering: AI output is not a substitute for approval from a competent human.

Common mistakes

  • Asking without observing. AI makes up; Feedback should be based on real events.
  • Attacking personality. Don't say "you are lazy", say "this behavior created that effect".
  • One-way feedback. Only the negative is unjust, only the positive is ineffective.
  • Submit the draft as is. Personalization and fairness control are a must.
  • Basing the result on AI. "AI scored like this" is neither ethically nor legally defensible.
  • Uncertain future step. Feedback should end with a concrete and agreed-upon next step.

In summary

  • Focus feedback on the behavior, not the person; Let the AI ​​structure the draft with a framework like SBI (Situation-Behavior-Impact) or STAR.
  • Content should always be based on your actual observations; Asking without observing will lead the AI ​​to make up.
  • Balance strengths and development areas; One-way feedback is either unfair or ineffective.
  • Prepare for difficult conversations and self-evaluation with AI, but the human always leads the conversation.
  • Do not forward the draft without reading it; Final evaluation, fairness approval and consequences (promotion/bonus) are the responsibility of the manager.

Application task

Write 4-5 concrete observations (date, event, impact, number if possible) from the last quarter for a (or imaginary) employee you manage. Have the AI ​​translate these into a balanced feedback draft within the SBI framework. Then have them produce "3 possible defensive reactions and responses." Finally, check the draft for fairness and realism: is each sentence based on an observation, are there strengths, is the next step clear? If there is even a single fabricated claim, remove it.

checklist

  • [ ] Is the feedback based on real, concrete observations?
  • [ ] Is the language focused on behavior rather than personality?
  • [ ] Are strengths and areas for development balanced?
  • [ ] Was a framework like SBI/STAR used?
  • [ ] Has the draft been personalized and checked for fairness?
  • [ ] Is there a concrete, understandable next step?
  • [ ] Did the final decision and responsibility remain with the human?
Previous Next