Logistyka
Wsparcie, które się nie wypala
Jak zdjęliśmy 75% obciążenia z dwóch operatorów i przywróciliśmy klientom szybkie odpowiedzi.
800 maili tygodniowo, 2 operatorów wypala się i nie nadąża. Klienci czekają godzinami.
A logistics company received 800+ email tickets per week: lost cargo, damage, delays, document requests.
Two operators worked 10-hour days and replied 4–6 hours late even to routine questions. Customers were churning to competitors.
Hiring another operator was hard: the niche labour market is thin, onboarding takes 3 months.
AI-agent klasyfikuje zgłoszenia i odpowiada szablonem, eskaluje trudne do człowieka. Oszczędność $5 400 / mies.
- 01
Analysed 4,000 past tickets and identified 12 main request types with template answers.
- 02
Built an AI agent on the Claude API with RAG over the contracts and template-answer knowledge base.
- 03
Wired up n8n for classification: 75% of tickets close automatically, complex ones escalate to humans with full context and a draft reply.
- 04
Added a quality-control dashboard: every auto-reply can be reviewed and rolled back.
Rezultat
75% of tickets close automatically in 30 seconds instead of 4 hours
Payroll savings: $5,400/month (one operator no longer needed)
ROI reached 1.5 months after launch
Customer NPS jumped from 31 to 58 in a quarter
Następny case
Od seed-rundy do produkcji
SaaS founder