Chatbot Development in Munich, Germany

Insurance / Personal Lines / Chatbot Development / Munich

A Munich-based insurance group (EUR 8.4B GWP, 9.2 million personal lines policyholders across Kfz/motor, Hausrat/contents, Haftpflicht/liability, Wohngebäude/buildings, and Rechtsschutz/legal expenses, with 4,800 exclusive agents and a 2,200-person contact centre in Munich and Nuremberg) needed a chatbot to handle claims notification and policy servicing — the two highest-volume customer interaction categories. Core challenges: (1) Schadenmeldung (claims notification) volume — 2.8 million annual claims across all lines. Kfz (motor): 1.4 million — ranging from Bagatellschäden (minor damage) to Totalschäden (total loss). Hausrat/Wohngebäude: 680,000 — water damage, storm damage, break-in, fire. Haftpflicht: 520,000 — third-party liability claims. Rechtsschutz: 200,000 — legal expense claims. Each claim requiring: Schadenmeldung (notification with details), Dokumentation (photos, police reports, estimates), and Schadenbearbeitung (processing). Contact centre handling 3.8 million claims-related calls (initial notification plus status inquiries and document submission). Average Schadenmeldung by phone: 14 minutes. Contact centre cost: EUR 42M annually. (2) Fachberatung expectation — German insurance customers expecting knowledgeable service. Kfz claims: customers asking "Muss ich nach einem Wildunfall die Polizei rufen?" (Must I call the police after hitting a deer?) expecting the chatbot to know: yes, in most Bundesländer this is required, the Jagdpächter (hunting tenant) must be notified, and Wildschäden (wildlife damage) are covered under Teilkasko, not Vollkasko — different deductible applies. Hausrat claims: "Ist der Wasserschaden durch die undichte Waschmaschine gedeckt?" (Is water damage from a leaking washing machine covered?) expecting: Leitungswasserschaden (pipe water damage) is covered under Hausrat, but whether a washing machine connection qualifies depends on the specific policy conditions (typically yes, as it's part of the Leitungswassersystem). The previous FAQ chatbot answering "Bitte prüfen Sie Ihre Versicherungsbedingungen" (Please check your insurance conditions) was rated 2.4/5.0 — customers feeling the chatbot added a step rather than helping. (3) Kfz-Wechselsaison (motor insurance switching season) — German motor insurance having a November 30 Kündigungsfrist (cancellation deadline) creating a massive autumn surge. October-November: claims and policy servicing volume increasing 2.5x as customers compare policies, submit changes, and process end-of-year claims. Contact centre deploying 400 temporary staff for 8 weeks — cost EUR 6.2M, training taking 3 weeks, and quality during Wechselsaison measurably lower than normal operations. (4) Betriebsrat requirements — the 2,200 contact centre employees represented by Betriebsrat. Previous technology deployments delayed 6+ months by Betriebsrat objections. Specific concerns: chatbot reducing headcount (threatening jobs), chatbot monitoring agent performance (surveillance), and chatbot quality being insufficient to represent the company (quality pride). The project needing Betriebsrat buy-in from day one. (5) BayLDA compliance — Munich's chatbot processing: customer identification data (name, address, policy number), claim details (potentially sensitive — health information in liability claims, security information in theft claims), and financial data (bank accounts for claim payments). BayLDA requiring: data minimisation (collecting only what's necessary for the claim), purpose limitation (claim data used only for claims processing), and specific consent for any processing beyond the primary purpose.

We built a Schadenmeldung and policy servicing chatbot over 10 weeks — handling claims notification with Fachberatung quality, seasonal surge management, and Betriebsrat-compatible design. (1) Intelligente Schadenmeldung (intelligent claims notification): conversational claims intake replacing structured forms. Natural language claims: "Gestern Abend ist mir ein Reh vors Auto gelaufen auf der B13 bei Dachau. Am rechten Kotflügel und der Stoßstange gibt es Schäden." (Yesterday evening a deer ran in front of my car on the B13 near Dachau. There's damage to the right fender and bumper.) → chatbot extracting: date/time (gestern Abend/yesterday evening), cause (Wildunfall/wildlife collision), location (B13 bei Dachau), damage (rechter Kotflügel, Stoßstange/right fender, bumper). Then providing Fachberatung: "Bei einem Wildunfall müssen Sie den Unfall bei der Polizei melden und den Jagdpächter informieren lassen. Wildschäden sind unter Ihrer Teilkaskoversicherung mit einer Selbstbeteiligung von 150€ gedeckt. Haben Sie bereits die Polizei verständigt?" (For a wildlife collision, you must report to police and have the hunting tenant informed. Wildlife damage is covered under your partial comprehensive with EUR 150 deductible. Have you already notified the police?) Coverage assessment: the chatbot checking the customer's specific policy and confirming coverage before the customer even asks — "Ihr Vertrag 12345678 deckt diesen Schaden unter Teilkasko ab. Ihre Selbstbeteiligung beträgt 150€." (Your policy 12345678 covers this damage under partial comprehensive. Your deductible is EUR 150.) Photo upload: the chatbot guiding photo documentation — "Bitte fotografieren Sie den Schaden: 1) Gesamtaufnahme des Fahrzeugs, 2) Nahaufnahme des beschädigten Kotflügels, 3) Nahaufnahme der Stoßstange" (Please photograph the damage: 1) Full vehicle shot, 2) Close-up of damaged fender, 3) Close-up of bumper). Photos processed with damage assessment AI providing preliminary severity estimate. Werkstatt (repair shop) coordination: chatbot offering partner workshops near the customer — "Die nächste Partnerwerkstatt ist Autohaus Müller in Dachau, 3km entfernt. Soll ich einen Termin für einen Kostenvoranschlag vereinbaren?" (The nearest partner workshop is Autohaus Müller in Dachau, 3km away. Shall I schedule an appointment for a cost estimate?) (2) Fachberatung-Qualität (expert advisory quality): the chatbot matching trained insurance agent knowledge. Coverage knowledge base: comprehensive German insurance coverage rules — understanding Teilkasko versus Vollkasko coverage scopes, Hausrat versus Wohngebäude coverage boundaries (Hausrat = moveable contents, Wohngebäude = building structure), Haftpflicht coverage limits and exclusions, and Rechtsschutz waiting periods and coverage areas. Scenario intelligence: "Mein Keller ist überflutet — was ist gedeckt?" (My basement is flooded — what's covered?) → chatbot determining: cause matters — Starkregen (heavy rain) through ground infiltration may be covered under Elementarschaden extension (if purchased), Leitungswasser (pipe water) is standard Wohngebäude coverage, and Rückstau (backflow) through drains may require specific endorsement. Asking clarifying questions to determine the correct coverage assessment. Regulatory accuracy: every coverage response caveated appropriately per German insurance regulation — "Diese Einschätzung basiert auf Ihren allgemeinen Versicherungsbedingungen. Die endgültige Deckungsentscheidung erfolgt durch unseren Schadenregulierer nach Prüfung aller Umstände." (This assessment is based on your general insurance conditions. The final coverage decision is made by our claims adjuster after reviewing all circumstances.) (3) Kfz-Wechselsaison management: seasonal surge handled by chatbot. Wechselsaison functions: during October-November, the chatbot handling: premium comparison ("Wie ändert sich meine Kfz-Prämie nächstes Jahr?"/How does my motor premium change next year?), coverage adjustment (SF-Klasse/no-claims class inquiry, Deckungserweiterung/coverage expansion), and Kündigungsfrist reminder ("Ihre Kündigungsfrist für den Kfz-Vertrag ist der 30. November. Möchten Sie Änderungen vornehmen oder haben Sie Fragen zu Ihrem aktuellen Vertrag?"/Your cancellation deadline is November 30. Would you like to make changes or have questions about your current policy?). Retention: when customers indicate intention to switch, the chatbot presenting: current coverage details, premium breakdown, claims-free bonus status, and loyalty benefits — information that helps customers make informed decisions rather than aggressive retention tactics. Volume handling: chatbot absorbing 60% of Wechselsaison inquiry volume — reducing temporary staffing from 400 to 150 agents. (4) Betriebsrat-kompatibles Design: designed with works council from concept stage. Agent augmentation, not replacement: the chatbot positioned as Mitarbeiterunterstützung (employee support) — handling routine inquiries so agents can focus on complex claims that require human judgement and empathy. Agent-facing tools: the chatbot also providing agents with real-time assistance — when an agent handles a complex call, a companion chatbot suggesting relevant policy details, coverage assessments, and next steps. This feature championed by agents themselves during design sessions. No individual monitoring: chatbot analytics track team-level and category-level metrics — not individual agent performance. Betriebsvereinbarung signed 3 weeks into the project — the fastest technology agreement in the company's history, attributed to early Betriebsrat involvement and agent-augmentation positioning. (5) BayLDA/DSGVO compliance: Datensparsamkeit (data minimisation) — chatbot collecting only information necessary for each specific interaction. Claims notification: damage details and photos required, health information only if injury claim, financial data only when payment processing needed. Transparent processing: customers informed at conversation start: "Ihre Angaben werden ausschließlich zur Bearbeitung Ihres Anliegens verwendet und gemäß unserer Datenschutzrichtlinie gespeichert." (Your information is used exclusively for processing your request and stored per our privacy policy.) Löschkonzept: conversation data retention periods per BayLDA guidance — claims data retained per legal requirement (Versicherungsvertragsgesetz), non-claims conversations deleted after 90 days unless customer requests earlier deletion.

Results over 12 months. Claims: chatbot handling 1.2 million Schadenmeldungen — 43% of total claims notifications. Phone claims calls from 3.8M to 2.4M — 37% reduction. Average Schadenmeldung: 6 minutes via chatbot (versus 14 minutes phone) — 57% faster. Damage photo submission through chatbot: 840,000 — structured photo guidance reducing incomplete documentation by 42%. Partner workshop appointments: 280,000 booked through chatbot — Werkstatt appointment within the claims conversation rather than separate interaction. Fachberatung: coverage assessment provided with 94% accuracy (validated against claims adjuster decisions) — customers receiving immediate coverage guidance rather than waiting for adjuster review. Customer satisfaction: chatbot-submitted claims satisfaction 4.3/5.0 (versus 3.8/5.0 phone) — customers appreciating instant coverage guidance and photo documentation capability. Kfz-Wechselsaison: chatbot absorbing 60% of seasonal surge. Temporary staffing from 400 to 150 — EUR 3.8M annual savings. Wechselsaison customer satisfaction from 3.1/5.0 to 4.0/5.0 — instant responses versus 35-minute phone waits during peak. Retention during Wechselsaison: 2.4 percentage points improvement — information-based retention through chatbot outperforming phone-based retention where agents were rushed. Betriebsrat: Betriebsvereinbarung signed in 3 weeks. Agent satisfaction: 78% positive response to chatbot — specifically valuing the agent-facing companion tool that provides real-time assistance during complex calls. Zero headcount reduction achieved through natural attrition redeployment — temporary staffing reduction did not affect permanent employees. Financial: contact centre reduction EUR 14.8M; Wechselsaison staffing EUR 3.8M; claims processing efficiency EUR 6.2M; workshop coordination EUR 2.4M; retention improvement EUR 8.8M. Total annual impact: EUR 36M. Development: EUR 620K. Annual platform: EUR 185K. ROI: 58x first-year return.