thinkhard.SI

Prompt repository Hiring & people

Turn engagement survey results into an action plan

You need a few priorities, owners and actions from survey results that managers can actually carry out.

Registry no.
PPL-015
Collection
Hiring & people
Laid out for
Any chat model
Score
10 out of 10
Blanks to fill
5
Added

Hiring & peopleAny chat

The prompt

Score 10 of 10, Excellent

Use it when You need a few priorities, owners and actions from survey results that managers can actually carry out.

  • {{audience}}
  • {{timeframe}}
  • {{survey_results}}
  • {{constraints}}
  • {{min_group_size}}
Full prompt and why it works
Prompt
Turn the engagement survey results below into an action plan for {{audience}}, covering {{timeframe}}. The plan should give managers a few priorities they can actually carry out, and show employees that their answers led to something.

Survey results (scores by question or theme, change since the last survey, response rate, and comment themes with counts):
"""
{{survey_results}}
"""

What is already in progress, and what limits us (budget, time, other changes under way):
"""
{{constraints}}
"""

Produce:
1. Priorities: no more than three. Choose them by three tests, and say which tests each one meets: a low or falling score, a theme many comments raise, and something within our control in this timeframe.
2. For each priority, a table with the columns Action, Owner, First step and date, How we will know it worked. Two to four actions per priority. Use owner roles only where the material names them; otherwise write [TO ASSIGN].
3. Not now: other issues the results raise that we will not act on yet, with one sentence each on why.
4. A message of under 150 words that managers can share with their teams: what we heard, what we will do, and when they will hear back.

Every action should be concrete enough that a manager could start it next week, for example "Each manager runs a 30-minute team discussion on the workload results and records the top two fixes" rather than "improve communication". The example shows the form only.

Rules:
- Use only the results above. Do not invent scores, benchmarks or causes. State a likely cause as a question to test with teams, not as a fact.
- Do not report on any group smaller than {{min_group_size}} people, do not try to work out who said what, and do not draw conclusions about any demographic group.
- Do not propose actions that need budget or headcount the constraints do not allow.

If the response rate is low, the results give no comparison, or numbers are missing, say so at the top and explain how it limits what the plan can claim. Do not guess at the missing figures.

Variables

{{audience}}
Who uses the plan, for example "the leadership team and 12 people managers".
{{timeframe}}
The period the plan covers, for example "the next two quarters".
{{survey_results}}
Scores by question or theme, changes since last time, the response rate and comment themes with counts. Remove anything that identifies a person.
{{constraints}}
Work already under way and limits on budget, time or people. Write "none" if there are none.
{{min_group_size}}
The smallest group you report on, to protect anonymity, for example 5.

Why it works

  • Length. No more than three priorities, chosen by stated tests, keeps the plan small enough for managers to carry out alongside their normal work.
  • What good looks like. Each action must be concrete enough to start next week, which rules out vague actions like "improve communication" that never happen.
  • What to avoid. A minimum group size and a ban on working out who said what protect the anonymity that makes people answer honestly next time.

How to use it

  1. Press Fill in variables and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.