Filled Examples
These short examples show the level of specificity expected in key artifacts. They do not replace templates; they demonstrate what a gate-ready entry looks like.
Initiative card
Title: Support operator assistant for standard requests
Process: first-line customer support
Problem: operator spends 8-12 minutes searching instructions and preparing an answer
Initiator: head of customer support
Business owner: contact center director
Impact hypothesis: if the operator gets an answer draft and source links in the interface, average handling time for standard requests will decrease by 20%
Baseline: AHT 9.5 minutes for selected request category
Target: AHT 7.6 minutes without CSAT degradation
AI product: knowledge assistant
Risk: employee hint; human sends the final answer
Next step: check knowledge base and select 50 standard requests for pilot
Process passport
Process: processing incoming supplier invoices
Owner: head of facilities
Volume: 2,000 invoices/month, ~8 min manual entry per invoice
Steps: read PDF → extract fields → accountant review → enter in ERP
AI fit: high (volume, digital data, text extraction, review exists)
Route: extraction — RAG/LLM; entry — agent
Risk tier: Moderate
Value hypothesis: 8 min → 2 min × 2,000 = ~200 h/month saved
Impact owner: head of facilities
Boundary: poor scan quality — confidence threshold and manual fallback
Use case document
User: first-line operator
Trigger: new customer request
Input: request text, category, available instructions
AI action: find relevant instructions and prepare answer draft
Human action: verify, edit if needed, and send to customer
Constraints: AI does not change customer data or decide compensation
Quality metric: share of answers accepted without major edits
Impact metric: AHT, FCR, CSAT
Experiment report
Period: 2 weeks
Group: 20 operators, 1 request category
Volume: 1,200 requests
Quality result: 78% of suggestions rated useful
Process result: AHT decreased from 9.5 to 8.1 minutes
Issues: in 12% of cases instructions were outdated
Decision: update knowledge base and expand pilot after content review
Impact report
Baseline: AHT 9.5 minutes
Target: 7.6 minutes
Fact: 7.9 minutes after 6 weeks in production
Additional metrics: CSAT did not decrease, FCR increased by 3 p.p.
Attribution: comparison with control request category without assistant
Decision: partially confirm impact, scale after knowledge base update