01
Frame it.
One page. Name the use case. Who owns it. The problem in plain English. The OKR it ladders to. Most use cases die here because no one wrote them down.
02
Score it.
Three axes: Business Impact, User Desirability, Technical Feasibility. 1–5 each. Score by gut, defend with data after.
03
Question it.
Two short questionnaires: how today works, what AI would change · what data you have, what's missing. Surfaces every blocker before you build.
★ WORKED EXAMPLE · ENGAGEMENT SURVEY SYNTHESIZER
The Canvas · one page, one decision.
★ GENERAL
USE CASE: Engagement Survey Synthesizer
DEPARTMENT: People & Culture
★ STAKEHOLDER
ROLE: Head of People Analytics
NAME: Priya Menon
GROUP: People Ops · Engagement squad
★ OBJECTIVE & KEY RESULTS · STRATEGIC FIT
Business objective: Cut the time from survey close → published action plan by 80%, so manager response is fresh, not stale.
KR 1
Time-to-themes < 48 hrs (was 3-4 weeks)
KR 2
90% of manager teams get a custom DM-ready summary
KR 3
Action-plan completion +25% YoY
KR 4
100% PII removed pre-LLM (CLEAR-checked)
Strategic fit: Aligns with the "manager-first" people strategy. Fastens the loop between sentiment data and action. Internal · Q3 priority.
BUSINESS IMPACT
4.3
of 5
USER DESIRABILITY
4.5
of 5
TECH FEASIBILITY
3.8
of 5
★ OVERALL
4.2
SHIP IT
★ PROBLEM TO SOLVE
Engagement surveys generate 5,000+ free-text comments per cycle. People Ops takes 3-4 weeks to read, code, and theme them. By the time managers get their team summary, the moment has passed and the data feels stale.
★ USE CASE DESCRIPTION
An AI synthesizer that (1) strips PII, (2) clusters comments into themes per team, (3) drafts a 1-page DM-ready summary per manager with top 3 themes + suggested questions for the next 1:1. People Ops reviews · manager receives within 48 hrs.
★ WHY NOW
LLMs now handle clustering + summarization with HR-policy alignment we couldn't do 18 months ago. Internal Slack pulse data shows managers want this. Legal already approved PII-stripping flow for the 2024 cycle.
★ STEP 2 · IMPACT ASSESSMENT
Score each axis · defend the number.
Strategic business impact
How does this align with executive strategy?5
Direct line to manager enablement + retention OKRs. CPO sponsor.
How does it generate business value?4
Faster sentiment-to-action loop. Estimated 200 People Ops hrs/quarter saved.
Business change management timeframe?4
~6 weeks for People Ops adoption. Manager-side adoption requires one demo cycle.
User desirability
Who are the key personas?5
People Ops analysts (primary) · people managers (secondary · receive summaries) · HRBPs (escalation).
How appealing is the value prop?5
Managers have asked for "what should I do about engagement?" for years. This is the answer.
Change resistance?3
Some People Ops team members protective of qualitative coding craft. Mitigated by keeping human-in-loop review.
Technical feasibility
Implementation + operational risks?3
PII leakage if pre-processing fails. False themes from over-fitting on small teams (< 5 people).
Safeguards available?4
Existing PII redaction pipeline · sample QA per cycle · min-team-size threshold of 7.
AI/LLM fit?5
Core LLM strength: clustering + summarization of unstructured text. Strong fit.
★ STEP 3 · USE CASE QUESTIONNAIRE
The questions that surface every blocker.
ITEM
QUESTION
YOUR ANSWER · Engagement Survey Synthesizer
1
Describe the current business practice. What might be gained with AI as part of the existing process?
Today: People Ops manually codes 5,000+ comments over 3-4 weeks. With AI: 48-hour turnaround, per-manager summaries, fresh sentiment routed to the right people in time to act.
2
What is the qualitative AND quantitative value AI provides?
Qualitative: managers feel heard, action plans match the actual signal. Quantitative: ~200 People Ops hrs/quarter saved · +25% YoY action-plan completion · 90%+ manager coverage.
3
What constraints exist currently?
PII handling per region (EU, US, APAC) · legal approval cycle · existing engagement-platform export format · People Ops capacity to QA outputs.
ITEM
DATA SYSTEMS & SCOPE
YOUR ANSWER
1
What data systems are used? Additional ones to leverage?
Today: Glint / CultureAmp export (CSV). To add: Slack DMs to managers (post-summary delivery), HRIS team-roster lookup for routing.
2
Historical data for ML?
8 quarters of past survey comments already coded by People Ops · gold-standard training data for theme accuracy benchmarking.
3
How is data measured / stored?
Quarterly CSV exports · stored encrypted in HRIS-secure folder · 90-day retention post-action-plan-close. Move through AI without touching raw PII.
★ ESTIMATED IMPACT · 12 MONTHS
What "ship it" looks like in numbers.
PEOPLE OPS HOURS
↓ 200 / qtr
From manual coding · redeployed to action work.
TIME-TO-THEMES
↓ 92%
3-4 weeks → 48 hours.
MANAGER COVERAGE
↑ 90%
From ~40% who got summaries to 90%+ getting team-specific ones.
PAYBACK
2 cycles
Build + roll out in Q1. Net positive by Q3 engagement cycle.
Take the canvas. Fork it for your use case.
Original 3-slide PowerPoint template · drop in your own use case · score each axis · take to your CPO.