20Finding & Prioritizing AI Use Cases
A practical method to discover AI opportunities, score them on impact, feasibility and risk, and pick a first pilot that's almost guaranteed to succeed.
What you'll learn
- Run a task inventory and process-mapping session
- Score use cases on impact, effort and risk
- Use an impact/effort matrix to prioritize
- Write a one-page pilot brief
Step 1 — Inventory the work
Interview the people doing the work (or yourself) and list recurring tasks. For each, capture:
| Field | Example |
|---|---|
| Task | Answer Messenger inquiries about rooms and rates |
| Trigger | New Messenger message |
| Frequency | ~60 per day, peaks evenings |
| Time per instance | 3–5 minutes |
| Inputs / systems | Messenger, rates sheet, booking calendar |
| Output | Reply; sometimes a booking |
| Pain | Slow nighttime replies lose bookings |
| Risk if wrong | Medium — wrong rate quoted |
Step 2 — Score each candidate
| Criterion | Question | Score 1–5 |
|---|---|---|
| Volume | How often does it happen? | 5 = many times daily |
| Time | How long does each instance take? | 5 = 15+ minutes |
| Value | Does it affect revenue or customers directly? | 5 = directly drives sales |
| AI fit | Is it language/document heavy with clear examples of good output? | 5 = very clear |
| Data readiness | Is the needed info available and accessible? | 5 = in one clean place |
| Risk (subtract) | What's the damage if AI is wrong? | 5 = legal/financial harm |
Priority score = Volume + Time + Value + AI fit + Data readiness − (Risk × 1.5). Anything above ~15 is a strong candidate.
Step 3 — Plot impact vs. effort
Quick winsHigh impact · low effortDo first
Big betsHigh impact · high effortPlan carefully
Fill-insLow impact · low effortDo when convenient
Money pitsLow impact · high effortAvoid
IMPACT ↑EFFORT →
Step 4 — Choose the first pilot
The ideal first pilot is a quick win that is:
- Visible — people will notice the improvement
- Measurable — clear before/after numbers
- Low-risk — a human reviews outputs, mistakes are cheap
- Contained — one process, one team, 2–4 weeks
Step 5 — Write a one-page pilot brief
PILOT: AI-drafted replies to room inquiries — Sunset Cove Resort
Problem 60 inquiries/day; average first reply 6.5 hours overnight;
owner estimates 20% of late inquiries book elsewhere.
Solution AI drafts replies from rates/policies sheet within 2 minutes;
front desk approves via Messenger inbox; urgent cases flagged.
Scope Messenger only, English + Tagalog, 4 weeks.
Success Median first reply < 15 min · 90% drafts approved with light
edits · inquiry→booking rate up vs. last 4 weeks.
Risks Wrong rate quoted → rates sheet is the only source; human
approval on every message during pilot.
Owner Ana (front desk lead) · Builder: Jae (freelancer)
Cost Setup ₱25,000 · tools ~₱1,500/month
Decision Week 4 review: scale / adjust / stop.Key takeaways
- Inventory recurring tasks with trigger, frequency, time, systems, pain and risk.
- Score on volume, time, value, AI fit and data readiness, minus risk.
- Start with a visible, measurable, low-risk quick win.
- A one-page pilot brief aligns everyone on scope, success metrics and the go/no-go decision.
Knowledge check
0 / 3Q1Which is the best first pilot?
Frequent, visible, measurable and low-risk with human review.
Q2In the matrix, high impact + low effort is…
Quick wins deliver value fast at low cost — do them first.
Q3Why include “Decision: scale / adjust / stop” in a brief?
Pilots should end in an explicit decision based on measured outcomes.