Case studies / Modena
AI-native customer support without the AI-startup budget
Modena’s support team handles inbound emails in three languages and serves two distinct audiences — borrowers and investors — under licensed-financial-institution audit requirements.
- Annual direct financial gain (5 agents)
- 27 420 €
- Time saved by AI features
- 27.5h / week
Challenge
Every customer reply at a licensed credit institution must be auditable and regulation-compliant. The support team was juggling three languages, two portals (modena.ee for lending and modena.capital for investors), and the constant pressure of fast, consistent quality across a small team.
Solution
- 11 AI workflows, one platform. AI is integrated into 11 distinct support workflows rather than bolted on as a side chatbot; every AI output goes through human review before reaching a customer.
- Cost-aware three-model strategy. Triage runs on Haiku, reply drafting on Sonnet, complex reasoning on Opus — admins can see and override the choice per scenario.
- RAG on the Zendesk archive. Every new ticket surfaces the three most similar resolved tickets as context, so the AI learns from how humans actually handled the same question.
- Portal-aware knowledge base. Every KB article is tagged with a portal — investor tickets never see lending FAQs and vice versa.
- Self-extending KB. Every Sunday, Claude reviews resolved tickets and drafts the three most-recurring FAQs; admins click ‘publish’ after review.
- Graceful degradation. If the Claude API is down, the platform keeps running — AI is an enhancement, never a critical path.
- AI
- Backend
- PHP
- Self-Service
- UX
Result
- Response time dropped from 5–10 minutes to 1–2 minutes per reply.
- Escalations now reach the right person the same day, not from the back of the queue.
- Reply quality is more consistent across agents — same KB, same prior resolutions.
- New hires use the platform on day one — no ‘prompt engineering’ training needed.
- Direct software savings of ~7 140€/year versus Zendesk Suite + Advanced AI.
Key takeaway
An AI-native support operation doesn’t need an AI-startup budget — it needs careful workflow integration, a cost-conscious model strategy, and humans in the loop on every decision.
