AI-Assisted Protection and Retirement Advice

How do scattered pieces of client information become an understandable advisory meeting? This interactive demo platform guides from a simulated CRM client lookup through preparation into thematic advisory journeys: advisor and client review existing contracts together, discuss coverage gaps and change assumptions directly in the interface. Zurich denotes the visible brand and product context of the demo, not a verified client relationship. Targeted AI text assistance complements the flow, while calculations and topic suggestions are based on traceable rules.

For the fictional persona Ennio, the advisory journey does not start with a rate card: the advisor first sees master data, life situation, policies and memos. Topic suggestions are created by stored keyword rules and make conversation triggers visible; they are not an autonomous assessment of the life situation. The interface distinguishes existing entries, suggested values and data gaps, and completion suggestions must be confirmed or corrected. New memos can be captured by voice and condensed by generative text assistance; they are stored locally in the browser.

For open items, the preparation provides a link with a QR code and an editable email draft that opens the local mail program. On the client side, entering selected open fields is demonstrated on a reduced interface. The boundary is part of the concept: transmitting the entries back to the advisor is not implemented in the current state. The screen shows the intended client participation, not an already closed data loop.

During the session, the prepared context turns into a concrete question: what should be protected or built up, and what is already in place? Income protection focuses on income needs, occupational context and existing benefits; retirement provision looks at the desired standard of living, current provision and a modeled pension gap together. Client data can be adjusted, contracts added and goals changed. The result is not a free-form AI forecast but a rule- and assumption-based calculation. The alignment with DIN 77230 structures the demo but does not constitute a certification.

A comparison becomes easier to understand when its premises can be discussed. In retirement provision, advisor and client review return assumptions and the allocation across provision layers; in survivors' protection, the need can be determined via a rule of thumb or individual building blocks. The investment journey connects goals and risk profile with a demo fund universe and a portfolio selection in its own section oriented towards section 34f. Sliders and visualizations are conversation tools: they imply neither a guaranteed return nor a binding price.

From the discussed need, the interface leads to suitable selection views: depending on the journey, product options, rate tiers or portfolio positions are reviewed. Selected insurance options come together in a shared selection with premiums and a before-and-after view. Recommendations are based on stored rules and demo product data; the premiums shown are not binding offers from a rating engine. The professional review stays with the advisor.

The protocol does not start with a blank page: the course of the session provides a rule-based pre-fill for occasion, wishes, recommendation, open points and next step. The advisor reviews and edits these five building blocks, and generative AI can work optional spoken corrections into the existing texts. A protocol PDF is generated from session data and a template. The subsequent handover shows a simulated process, not a verified production application submission, and the stored follow-up appointment does not replace an external calendar booking.

The application visibly distinguishes which content was calculated, suggested or formulated: the memo and protocol assistants use generative AI via server-side calls, while topic suggestions, needs models and selection rules remain deliberately rule-based. The investment copilot and the reading of policies are simulated in the views examined. Assistance is thus a tool of advice, not an autonomously acting team of agents: review remains part of personal advice.

The web application combines a React and TypeScript frontend with an Express backend, and session data is stored in PostgreSQL via APIs. Shared calculation and rule modules keep specialist derivations separate from the presentation. Generative text functions for memo preparation and the protocol run through server-side OpenAI calls, separate from the mock copilot, demo pricing and rule-based calculations. For classifying the application, this separation matters more than the blanket label of an AI platform.