Customer support · Production

Tryg deploys three conversational AI virtual agents across Norway, Denmark, and Sweden for customer and internal support

The problem

Insurance complexity left Tryg's customers with unanswered questions about policy wording and product information across three distinct markets, while the company also needed to improve the efficiency of its support staff handling high inbound query volume.

Workflow diagram · grounded in source
1
Customer or employee initiates chat
trigger
“a conversational AI-powered virtual agent that specializes in helping customers with insurance claims”
2
Natural language processing of query
ai_action
“Using natural language understanding, the virtual agent can also identify and recommend solutions to customers for products that they may be eligible for to expand their coverage”
3
Autonomous resolution or escalation
routing
“The more Mia can assist customers on her own, the more it frees up the customer service teams to tackle complex customer queries”
4
API-driven self-service tasks
integration
“utilizes API integrations to automate a variety of tasks related to pricing, policy coverage and status, and more”
5
AI Trainers maintain knowledge base
feedback_loop
“The AI Trainers responsible for Rosa not only keep her knowledge up-to-date so that she can assist with relevant answers”
Reported outcome

Mia automated 80% of customer cases with over 200,000 conversations in 2020; Rosa answered correctly in 95% of cases while logging more sessions than calls to the back office; and Ebbe was deployed to the Swedish market in just 3 weeks despite an original six-month timeline.

Reported metrics
Mia automation rate80%
Mia conversations in 2020over 200,000
Mia topics covered5,000
Rosa accuracy rate95%
Show all 9 reported metrics
Mia automation rate80%
Mia conversations in 2020over 200,000
Mia topics covered5,000
Rosa accuracy rate95%
Rosa employees assisted daily750
Rosa topics covered1,200+
Rosa sessions vs back office callsmore monthly sessions with Rosa than calls to the back office
Ebbe deployment time3 weeks
Ebbe original planned deployment timelinesix-month timeline
Reported stack
boost.ainatural language understanding
Source
https://www.boost.ai/case-studies/tryg-case-study-conversational-ai
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Mia automated 80% of customer cases with over 200,000 conversations in 2020; Rosa answered correctly in 95% of cases while logging more sessions than calls to the back office; and Ebbe was deployed to the Swedish mark…

What tools did this team use?

boost.ai, natural language understanding.

What results were reported?

Mia automation rate: 80%; Mia conversations in 2020: over 200,000; Mia topics covered: 5,000; Rosa accuracy rate: 95% (source-reported, not independently verified).

How is this customer support AI workflow structured?

Customer or employee initiates chat → Natural language processing of query → Autonomous resolution or escalation → API-driven self-service tasks → AI Trainers maintain knowledge base.