Human ResourcesEmployee Experience

AI Prompts for Employee Retention Analysis

Retention analysis starts by measuring the right thing: total attrition mixes retirements, restructuring and poor performers with the departures that hurt — regretted leavers, high performers, people in scarce roles. It continues by finding where and when attrition concentrates — role families, tenure bands, teams, seasons — and what changed for those groups. It ends with actions aimed at causes.

These prompts compute the metrics, analyze concentrations, and build the actions. The guardrail matters especially here: this is group-level analysis to improve conditions and support, not a tool for predicting or flagging which individuals will leave.

Before you use these

Have these ready to replace the highlighted [variables]:

Headcount and leaver data by period, with role family, level, tenure, team (above threshold), and leaver type (voluntary/involuntary, regretted/not)
Definitions your organization uses for regretted attrition
Context: reorganizations, pay reviews, market changes, manager changes
Cost assumptions for replacement if you want a cost view

These prompts analyze workforce data in aggregate to improve policy, management and support. They must not be used to score, rank or flag individual employees, to predict which named person will leave, or to inform decisions about individuals; analysis is reported at group level above the anonymity threshold, and any group-level finding is a prompt for a conversation, not a conclusion about people. Check local law and policy on workforce data use.

The prompts

1. Compute attrition metrics that mean something
2. Analyze where and when attrition concentrates
3. Build retention actions aimed at causes

1. Compute attrition metrics that mean something

Best forAttrition rates by type and group, computed consistently, with the definitions stated.

Inputs needed

Headcount and leaver data
Definitions
Period

How to use itPaste the data. The model computes annualized attrition by type and group, regretted attrition, first-year attrition, and internal mobility, with formulas and the anonymity threshold applied.

Expected outputMetric table with formulas: total, voluntary, regretted, first-year, by group; internal mobility; trend; and the definition notes.

Act as a people analytics partner computing attrition metrics for [company / period].

Data: [average headcount and leavers by period; leaver type (voluntary/involuntary; regretted/not regretted); role family, level, tenure band, team above threshold]
Definitions: [regretted attrition as we define it; involuntary categories]
Anonymity threshold: [n]

1. Formulas stated: attrition rate = leavers ÷ average headcount, annualized; voluntary; regretted (and the definition used); first-year attrition (leavers within 12 months of hire ÷ hires in the cohort); internal mobility rate.
2. Metrics overall and by group above the threshold; trend versus prior periods.
3. Regretted versus unregretted: the share, and where regretted attrition concentrates.
4. First-year attrition as a hiring/onboarding signal, by role family and cohort.
5. Cost view (if assumptions given): replacement cost × regretted leavers, with the assumptions explicit.
6. Data quality: missing leaver types, inconsistent recording, and the effect on the metrics.

Show the arithmetic. Report only above the threshold; suppress cells that would identify individuals.

2. Analyze where and when attrition concentrates

Best forFinding the groups, tenure bands and periods where regretted attrition is unusually high, and what changed for them.

Inputs needed

Metrics by group and period
Context events
Exit and survey findings

How to use itThe model compares groups to the organization baseline, checks whether differences are large enough to be meaningful, aligns spikes with context events, and cross-references exit and survey data — at group level.

Expected outputConcentration findings with baseline comparison and materiality, timing patterns with aligned events, corroboration from exit and survey data, candidate causes per concentration, and what needs more evidence.

You are analyzing attrition concentrations for [company / period] at group level.

Metrics: [regretted and voluntary attrition by role family, level, tenure band, team above threshold, by period]
Context: [reorganizations, pay reviews, manager changes, market events, with dates]
Exit and survey findings: [coded reasons by group; engagement drivers by group]

1. Concentrations: groups whose regretted attrition is materially above the baseline — state the rate, the baseline, the number of leavers, and whether the difference is large enough to trust given the group size.
2. Timing: periods with spikes, aligned with context events; tenure bands where attrition peaks (e.g. months 6–12, or post-promotion cycle).
3. Corroboration: for each concentration, what exit reasons and survey drivers say about that group.
4. Candidate causes per concentration (manager capability, growth ceiling, pay compression, workload, role design, change fatigue) with the evidence for each and the evidence that would confirm it.
5. Where the data is too thin: what to collect (stay interviews, a targeted pulse) before acting.
6. One-page summary: the three concentrations that matter most, the likely cause of each, and the confidence.

Group level only. Do not identify, score or flag individuals, and do not attempt to predict which named employees will leave.

3. Build retention actions aimed at causes

Best forActions targeted at the causes behind each concentration, with owners and the measure that shows they worked.

Inputs needed

Concentrations and candidate causes
Constraints
Leaders' commitments

How to use itThe model matches each cause to an action (manager support, career paths, pay review, role redesign, workload), estimates effect, assigns owners, and sets the measure and review date.

Expected outputAction plan by concentration with cause, action, owner, timing, cost, expected effect and measure; stay-conversation guidance for managers; the review cadence.

Act as an HR leader building retention actions for [company] from the concentration analysis.

Concentrations and causes: [group, rate vs baseline, likely cause, confidence]
Constraints: [budget, pay review timing, manager capacity]
Commitments: [what leaders have agreed]

1. For each concentration: the action that addresses the cause (not the symptom) — e.g. manager coaching and support for a team with a new manager; visible career paths for a plateaued role family; a pay-equity review where compression is evidenced; workload rebalancing; role redesign — with owner, timing, cost, expected effect on regretted attrition and the measure.
2. Stay conversations: guidance for managers to talk with their teams about what keeps them and what would make them leave — voluntary, framed as improving the team, not as identifying flight risks; the questions; what is fed back in aggregate.
3. Sequencing: quick actions versus structural ones; what to communicate and to whom.
4. Measures and review: regretted attrition by group at [n] months, the leading indicators (survey items, internal mobility), and the review date.
5. What not to do: retention bonuses without addressing the cause; singling out individuals; promises that cannot be kept.
6. Summary for leadership: three actions, owners, dates.

Aim at causes. Actions that treat symptoms delay the departures by a quarter.

Attrition metrics, computed

Illustrative figures showing why total attrition hides the number that matters.

Average headcount (year) = 420 Leavers = 71 → total attrition 16.9% involuntary (restructure, performance) = 19 voluntary = 52 → voluntary 12.4% not regretted (per definition) = 21 regretted = 31 → regretted 7.4% Regretted attrition by tenure band (n above threshold only): < 1 yr : 9 of 60 hires = 15.0% ← first-year problem 1–3 yrs : 14 of 150 = 9.3% 3+ yrs : 8 of 210 = 3.8% Regretted attrition by role family: Engineering 11 / 120 = 9.2% Customer ops 13 / 90 = 14.4% ← concentration; exit reasons: 'no path', 'manager' Other 7 / 210 = 3.3% Headline: 'attrition is 17%' → actionable: 'regretted attrition is 7.4%, concentrated in first-year hires and customer ops, with growth and management cited.'

Reported at group level above the threshold. The next step is an investigation of customer-ops conditions, not a list of names.

Related prompts

Logical next step

After this, most HR teams move on to Exit Interview Analysis.

Get the free HR & Recruiting AI Starter Kit → Nine of these prompts as an intake → assess → offer hiring workflow with an intake worksheet, delivered by email. See what's inside

All Human Resources prompts · Search the full library