United KingdomFinance and insurance2026 – ongoingMonzo Bank
LLM identification and escalation of vulnerable customers
Monzo said AI referred customers in crisis to safeguarding specialists more than 10 times faster
Problem
Monzo's support teams had to identify customers facing immediate harm among large volumes of chats. Customers expressed distress indirectly and in varied language, making identification difficult.
Approach
Monzo built an LLM system to analyse incoming chats and automatically escalate urgent cases to safeguarding specialists. Specialists helped label evaluation data and review ambiguous examples. The bank tested predictions against live conversations without triggering referrals, then enabled automatic escalation gradually with rollout controls. Staff retained manual escalation for cases the model missed.
Results
Takeaway
Testing against live traffic helped Monzo assess referral volumes before exposing specialists to automatic escalations. Manual escalation remained necessary because the model could miss vulnerable customers.
Categories
- Industry
- Finance and insurance
- Business function
- Customer serviceFinance, risk and legal
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Monzo described its deployed LLM-based safeguarding referral system and staged rollout. It reported referrals within seconds, more than 10 times faster than before, and explained its ongoing specialist review and monitoring.
Full details are in the original sources.