Conversational AI & Customer Dialogue

Included

• Enquiry classification by complexity and AI suitability
• Tooling concept: classification, draft responses, FAQ bot
• Vendor selection and Swiss-nDSG data protection impact assessment
• Training on prompt adjustment and escalation paths

How we work

1. Enquiry inventory: we sort the real ticket volume by frequency, effort and escalation risk
2. Drawing the line: which enquiry types suit automation, which must stay with people
3. Tooling concept: automatic classification, LLM draft responses with human approval, FAQ bot for standard questions
4. Vendor selection and nDSG impact assessment for customer data in prompts
5. Rollout with the service team: prompt adjustment, escalation paths, quality measurement

Examples from our practice

• Swiss service provider with high enquiry volume: classification of inbound mail, draft responses with human approval for medium complexity, FAQ bot for standard questions. The goal was explicitly relief, not replacement.

What you concretely get

• Sorted enquiry inventory with an automation recommendation per type
• Tooling concept with a clear human/machine boundary
• nDSG assessment for customer data in prompts
• Handover concept and escalation paths the team maintains without us

Engagement model

6-10 weeks depending on enquiry volume. Effort 8-12 consulting days. Requires an anonymised export from the ticket system.