Insurance · South Korea
Korean insurance chatbot & voice assistant
Customers get answers from the insurer’s own policy documents, by chat or by phone, instead of generic chatbot guesses.
Result: Korean end to end · chat + phone
View case studyIndustry guide
Assistants and research agents that answer from your own policies, procedures and data, with citations, and that run privately inside your environment.
The problem
General chatbots describe what insurers usually do, not what your policy actually says. In a regulated business, that’s a risk.
Policy wordings, procedures, product sheets and past cases live in different systems. Staff spend hours finding what the company already knows.
Customer data and proprietary research questions can’t be sent to an outside API.
What we build
Every build is scoped against a metric you can rerun yourself, and you own the code and models.
RAG over your own documents that answers with citations, so staff and customers get the company’s actual rules, not generic advice.
The same grounded engine behind a phone line: speech-to-text, policy check, spoken answer.
Specialist agents that collect, check and score data into a structured report, with every number traceable to its source.
Open-weight models on your infrastructure when data can’t leave, with an honest view of the quality trade-offs.
Relevant work
Our Korean-language insurance assistant was benchmarked head-to-head against ChatGPT on the insurer’s own customer questions. The same method works for any regulated document set, in English or Korean.
Insurance · South Korea
Customers get answers from the insurer’s own policy documents, by chat or by phone, instead of generic chatbot guesses.
Result: Korean end to end · chat + phone
View case studyFinance · investment research
Eight AI agents draft first-pass company research, with every score traceable to its inputs. Data stays in-house.
Result: 8 agents · 50+ metrics per company
View case studyTelecom · home networking
Vague “my internet is slow” tickets become a ranked root cause, the log lines that prove it, and the next action.
Result: 7 analysis passes · 3 vector memories
View case studyEngagement options
Scope first, build only if the numbers say it’s worth it, and add Operate if you’d rather we run it.
Fixed price · 1 to 2 weeks
We study your data and workflow, agree on the metric that defines success, and give you a fixed quote for the build.
Fixed price · prototype in 2 to 6 weeks, then production
A working prototype on your real data, then hardening, deployment and handover to your team.
Monthly · cancel with 30 days’ notice
We keep your system accurate, current and cheap to run, so your team doesn’t have to.
Common questions
Answers are generated only from retrieved passages of your documents and shown with citations. We measure accuracy on a golden set of real questions, and the system says “I don’t know” when the documents don’t cover a question.
Yes. When data can’t leave your network we deploy open-weight models on your hardware, and we design the pipeline around that constraint from day one.
Yes. We work in English and Korean, including on mixed-language document sets. Our insurance case study was in Korean; the same pipeline works in English.
Start a project
Two or three sentences about the data you have and the outcome you want is enough. We reply within 1 business day with an honest read on feasibility. No slide deck needed.