DField SolutionsMérnöki stúdió · Budapest
Loading · Töltődik
Skip to content
CASE STUDIES · fintech

AI support copilot for an EU neobank · 42% cost drop, 18% CSAT lift

A fast-growing EU neobank's support team was drowning in Tier-1 tickets. We built a retrieval-augmented copilot with strict guardrails, live traffic in week four, metered against real CSAT.

Timeline9 weeks
← Back to case studies
Reviewed by
01 / Cost per resolved ticket−42%
02 / CSAT+18 points on AI-handled conversations
03 / Tier-1 deflection61%
04Zero PII leaks during 12-week audit
The problem01 / 03
  • 01Tier-1 ticket volume grew 3× in 8 months; hiring couldn't keep up.
  • 02Previous vendor's chatbot hallucinated account balances · legal pulled it.
  • 03No eval pipeline; every release was a gamble.
  • 04Compliance required PII to never leave the EU VPC.
The solution02 / 03
  • 01Hybrid retrieval (pgvector + BM25 + reranker) over 12k KB articles.
  • 02LLM router: small model on FAQs, GPT-4o on reasoning. 3.4× cost cut.
  • 03Guardrails: PII scrubber, prompt-injection detector, refuse-to-answer.
  • 04CI eval suite with 180 golden questions; blocks deploy if factual accuracy < 95%.
  • 05Full observability: tokens, latency, quality, cost per tenant.
The outcome03 / 03
  • 01Cost per resolved ticket: −42%
  • 02CSAT: +18 points on AI-handled conversations
  • 03Tier-1 deflection: 61%
  • 04Zero PII leaks during 12-week audit
  • 05Live in week 4, full rollout by week 9
CASE STUDIES

Let's get started.

Send an email or book a 30-minute call.

Back to case studies