All business cases
Google Chat Retail

Supermarkets — Chatbot

A Google Chat agent guides employees through property damage and liability claims, GDPR-compliant, with near 100% file completion.

Case summary

A major European supermarket chain wanted to modernize claims intake for property damage and general liability (GL) across its stores. Until then, teams relied on paper forms or Excel files and had to follow different processes depending on whether the claim concerned damage to the store — broken glass, fire, theft — or a customer who was injured, such as a fall or injury. Collecting information was time-consuming, and incomplete files led to follow-ups and delays.

The retailer chose to deploy a conversational agent in Google Chat. From that chat, an employee starts the declaration by specifying the type of claim: store damage or customer injury/GL. The agent then guides the user through a dynamic flow, asking only the relevant questions.

Key information collected

  • Incident location: area — store, mall, parking, gas station, drive-through — exact store name, and the aisle or department if applicable.
  • Nature of the claim: injuries and/or material damage, cause and circumstances.
  • Date and time of the incident and the reporter's contact details.
  • Presence of witnesses, description of injuries if applicable, and victim information — identity, contact details, address for GL cases.
  • For material damage: estimated amounts, business interruption, and third-party liability.

The project was conducted with increased attention to GDPR. Field definitions, hosting, and data retention periods were aligned with the recommendations of the data protection officer and legal counsel, collecting only relevant information.

Fields become mandatory or disappear depending on context: the precise location is requested only after selecting the area, and questions about a liable third party appear only if a third party is involved. Collection of photos and documents is integrated; the agent reminds users that no individuals must appear in the images.

Solution and implementation

  • Single channel: Google Chat

    Employees start the declaration from the company's Google Chat app. The chatbot verifies their identity and offers either a property-damage claim or a liability claim.

  • Guided, conditional flows

    Flows branch depending on claim type. Property damage covers the building, goods, and equipment; GL claims collect circumstances and injury location. Only relevant questions are shown.

  • Validation and integration

    Codified dropdown lists ensure consistent answers. The agent generates a structured JSON file per case, sent securely to the broker or insurer's system.

~100% completion rate
Guided questioning and validations make declarations far more complete than before, eliminating most follow-ups.

Expected results

  • Complete files on first submission: thanks to guided questioning and validations, declarations are far more complete than before, eliminating follow-ups.
  • Reduced declaration time: a complete declaration takes only a few minutes, versus multiple calls and manual entries previously. Employees no longer need to log into an external portal.
  • Standardized data: normalized incident and circumstance codes enable statistical analysis and automated processing, with no re-keying.

Learn more

Production required building two main business guides.

Guide 1

Property Damage

For broken glass, fires, water damage, theft, storms and other material losses. This guide checks the area and location of the claim, collects estimated amounts, and asks whether a liable third party or business interruption is involved. If a third party is mentioned, their contact details are requested.

Guide 2

General Liability

Used when a customer or third party is the victim of bodily injury or material damage. It captures the circumstances, injury location — head, arms, legs — and severity, then the identity and contact details of the victim and witnesses. The agent suggests adding photos, but reminds users to keep images anonymous.

Internal classifications

Incidents are grouped by codes, facilitating reporting and case prioritization. Model fields are aligned with the partner insurer's standards so that transmission requires no adaptation.

Sample classification codes
D14 — Broken glass R01 — Slip and fall in store

To ensure personal data protection, only strictly necessary information is collected, and access to cases is restricted. The system automatically masks any non-relevant health data and deletes documents after processing.