How JanBask helped a 12-office insurance agency eliminate 65% of repetitive support tickets, cut response times by 40%, and scale operations without adding a single new hire — through a purpose-built AI support automation platform trained on their own workflows.
This regional property & casualty insurer managed policy inquiries, claims support, and billing questions across 12 offices — supported by a team of just 6 agents. As customer volume climbed, that lean team found itself buried in the same questions day after day.
Average response times had crept above 18 hours during peak periods. Customer satisfaction scores were slipping, and leadership faced a difficult choice: hire more agents, or find a smarter way to scale.
The business needed AI customer service automation that understood its actual workflows — not a generic chatbot dropped on top of a broken process.
JanBask designed and deployed a fully integrated AI support platform — not an off-the-shelf tool, but a purpose-built system trained on the client's own policy documents, claims procedures, FAQs, and 4,200+ real support tickets. AI workflow automation handled the repetitive queue while human agents stayed focused on the complex cases that needed real expertise.
A chatbot trained on 4,200+ real support tickets — not a generic template. It speaks the language of insurance, understanding terms like "fender bender" vs "total loss" and routing each accordingly.
Semantic search across policy documents and FAQs stored in Confluence — surfacing accurate answers in seconds rather than making customers wait 18+ hours for a human response.
Real-time classification of customer intent — automatically escalating urgent claims or complex cases to human agents while the AI handles routine queries without friction.
Tickets automatically assigned by topic, urgency, and branch location — eliminating the manual triage that was consuming hours of agent time across all 12 offices.
Bi-directional sync with the client's Salesforce CRM and Zendesk ticketing system. Every AI-resolved interaction updates the customer record automatically — no rip-and-replace required.
Real-time reporting on resolution rates, CSAT, and ticket volume across all 12 offices — giving leadership the performance visibility they never had before.
A structured delivery methodology that minimises disruption while ensuring the solution reflects how the business actually operates — not a generic playbook dropped on top of real workflows.
We audited 6 months of support tickets across all 12 locations, mapped inquiry patterns, and pinpointed the top 40 query types that represented 78% of total volume. This gave us a precise automation roadmap with no guesswork.
Our team built a generative AI solution trained on the client's policy documentation, internal FAQs from Confluence, and 4,200 historical Zendesk tickets. A key part of this phase was language calibration — the AI needed to learn that "fender bender" triggers a minor collision claims flow, while "total loss" requires a different escalation path with an adjuster. Those distinctions don't exist in generic models. The intent engine was integrated directly into Salesforce so every AI-resolved interaction updated the customer record automatically.
We launched with a soft rollout across 3 locations first, using live performance data to refine the model before expanding company-wide. Weekly optimisation sprints over the first 60 days continuously improved resolution accuracy.
All metrics tracked against pre-implementation baselines across all 12 locations. The full solution went live in week 10 of the engagement.
A production-grade AI stack chosen for reliability, scalability, and deep integration capability. From LLM-powered conversation handling to CRM synchronisation and cloud infrastructure, every component was selected to support secure insurance workflows and continuous optimisation as volumes grow.
LLM-agnostic architecture — the platform is not locked to GPT-4. The abstraction layer allows model swaps as better options emerge without rearchitecting the integration.
SOC 2-aligned data handling — customer PII never persists in the AI layer; conversation context is session-scoped and purged on close.
Human-in-the-loop escalation — agents receive full conversation context on handoff, so they never start from scratch on escalated cases.
We were at a breaking point — either hire more staff or find a smarter way to handle the volume. JanBask didn't just build us a chatbot. They understood our policy workflows and built something our agents actually trust. Handling 65% more routine volume without a single new hire is something we didn't think was possible.
© 2026 Copyright - JanBask.com | Designed by - JanBask Digital Design
