Vibeleaderboard
← All Intel
Intel / article

Webinar Recap How One Of Europes Largest Insurers Got Ai Agents Into Production

Source
elevenlabs.io
Author
ElevenLabs editorial sitemap
Date

Context

In ElevenLabs' webinar recap, insurer Admiral describes separating conversational judgment from required business checks. Admiral uses AI to interpret callers' intentions, while anything the business must be able to prove runs as fixed logic. ElevenLabs' Anna Neely describes workflow gates, such as authentication, that make functions available only when a condition is met. The account says vulnerable or distressed callers are routed to people.

Admiral also reports testing each change in simulated conversations before starting deployment at 1% of traffic. Expansion depends on customer-satisfaction and resolution metrics holding up. This describes a process for controlling changes after launch, with customer feedback informing whether deployment expands. Deployment account.

Key takeaways · AI-distilled
  • Admiral's first production voice agent handled settlement quotes in its UK lending business, picked because the conversation has few paths yet still exercises telephony and back-end integration, and was limited to one country and one business line.
  • Legal and compliance joined on day zero with a pilot scoped to a defined number of calls, and Admiral rebuilt its governance cadence because, per its team, a six-month approval cycle risks approving technology that is already outdated.
  • Admiral splits logic by provability: non-deterministic reasoning reads intent and detects distress, while required regulatory statements and mandatory paths run deterministically, and vulnerable or distressed callers are routed to humans.
  • Every agent change runs as a branch against hundreds to thousands of simulated conversations, then rolls out from 1% to 100% of traffic while CSAT and resolution are watched, a cycle Admiral says has shrunk from weeks to hours.
  • Prompts written in English underperformed in France, Spain and Italy; rewriting them natively in each market's language, rather than translating, produced responses that matched local customer expectations, according to the webinar recap.
Terms in this piece · Glossary
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Read the source elevenlabs.io
More from ElevenLabs editorial sitemap
Recommended reads
Comments

Checking sign-in…

Loading comments…