Usable HR feedback

How can you identify trends in anonymised HR feedback without erasing disagreements or exposing its authors?

1250 € per session, excl. VAT

In short

In a half-day session of 3 hours participants use generative AI to group anonymised HR comments, write a traceable summary and check its faithfulness. They leave with an operational note distinguishing trends, minority views and limitations, to be validated before any distribution.

Why this training?

Free-text comments contain nuances that overly broad categories can hide. A useful summary must preserve disagreements, distinguish frequency from importance, and avoid turning a few accounts into a general opinion. AI can suggest groupings and wording, but can also invent an explanation or downplay an isolated signal. The work therefore focuses as much on writing as on checking sources and protecting authors. Running case study: The fictitious organisation Virelune Cartonnages wants to understand feedback on its new induction programme. A training corpus of 36 fictitious comments combines positive assessments, practical difficulties, contradictions and ambiguous wording. The HR manager expects a collective note, without ranking individuals or comparing small teams. Transfer to the workplace: During the next internal survey, reuse the prompt and checklist on an authorised corpus, after checking its anonymisation. Compare the summary with the source comments, record corrections and obtain validation from a qualified HR professional before distribution.

Objectives of this training

  • Identify clues that could enable re-identification before submitting comments.
  • Formulate a grouping prompt with categories and source references.
  • Distinguish trends, contradictions and minority signals without extrapolating.
  • Write a note separating findings, limitations and avenues for discussion.
  • Measure the faithfulness of the summary by comparing it with the corpus.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • HR officers using internal surveys or induction questionnaires.
  • HR managers preparing collective feedback reports from free-text comments.

Prior skills

  • Have already made simple requests to a generative AI assistant.
  • Be familiar with the uses of internal surveys and your organisation's confidentiality rules.

What you will work on

  1. Secure the exercise corpus

    Objective: Define which information can be used without exposing authors. Direct identifiers, contextual clues and risks associated with small groups; the difference between anonymisation and simply replacing names.

  2. Group without forcing the meaning

    Objective: Build themes that remain faithful to the comments collected. Prompt specifying context, format, no invention and handling of ambiguities; possibility of several themes per comment.

  3. Write a balanced feedback report

    Objective: Produce a readable note without overinterpretation. Counting comments rather than people; distinction between occurrence, severity and representativeness; quotations and re-identification risks.

  4. Check and prepare for validation

    Objective: Assess faithfulness before sharing. Checking sources, omissions, counts and unwarranted wording; role of the qualified HR professional and limitations of automated assistance.

What you take away

  • Reusable summarisation prompt with explicit limitations.
  • Thematic table linked to comment identifiers.
  • One-page collective feedback note.
  • Completed quality checklist and correction log.

Method and assessment

  • Short explanations followed by immediate practice on the running case study.
  • Individual work followed by work in pairs, with peer review of outputs.
  • Guided group discussion: what works, what needs correcting and what remains a human decision.
  • Preparation: Have access to a generative AI tool authorised by the organisation. Identify the usual recipient of HR feedback reports; no real comments are needed for the session.

Observable criteria: No identifying clue detected during the exercise remains in the submitted data. ; Every finding in the note refers to verifiable comments. ; Counts specify their unit and match the thematic table. ; Identified contradictions and minority signals remain visible. ; The final check identifies omissions and unsupported assertions, then documents their correction..

Frequently asked questions

Can real feedback be used?

The exercises use the fictitious case study. Subsequent use requires an anonymised corpus and an authorised environment.

Should isolated views be removed?

No. Their low frequency must be explicit, without minimising their potential importance or identifying their author.

Can the note support an individual decision?

No. No sensitive decision is automated; the collective feedback report requires validation by a qualified professional.

What retention rules should be applied?

Follow internal rules; applicable obligations should be checked with your legal department.

Last updated: