InsurTech Spotlight: Zelros


Company: Zelros


Business Category: A provider of insurance-process services

Maturity Status: Mature (more than 3 years since product went into production)

Product or Service: Zelros platform

Area: Both Property & Casualty insurance and Life & Health insurance

Zelros is the industry’s first AI-driven platform dedicated to advancing insurance distribution. It enables incumbent industry leaders to compete with both fast-growth, technology-first insurance startups and tech behemoths heavily investing in InsurTech.

In a nutshell, it gives sales and marketing teams the opportunity to be more proactive, by equipping insurance with complete customer and policy knowledge to find the right policy at the right time. This technology leads to higher customer satisfaction and retention, lower attrition, and higher sales – both on digital and direct (human) channels (cross-sell and up sales).

What makes us an InsurTech:

  • Personalized advice: Zelros detects and predicts relevant customer situations to recommend best offers, advice or knowledge. All advice is based on insurance data mixed with Zelros’ proprietary insurance data catalog, providing insights from 3rd party data providers and open data.
  • Enterprise AI: Zelros leverages advanced analytics to build machine learning models for insurance. All models are documented in reports, ensuring a responsible use as well as full auditability of AI. All models are fully monitored to improve performance over time.
  • Data processing: Zelros provides standard data connectors to connect all insurance data whatever the format or source (datalake or business application). All data processing can be done in the secured Zelros cloud or on premise.
  • Deployment: Zelros can be deployed as a Salesforce App or as a fully configurable widget in any CRM or portal. Zelros is an API first platform that can be integrated in any application.
  • Security: Zelros will be certified ISO 27001 / 27019 in Q1 2022. All infrastructure is hosted on the highest level of security on Microsoft Azure cloud.

Major features & technology used:

  • Reinforcement Learning: in the Zelros context, an algorithm that provides a set of relevant selling points given the customer profile and automatically and instantly adjusts its predictions based on any source of feedback (thumb up / thumb down, quote creation, etc.)
  • NLP: it is a broad field that embeds all the technologies used to deal with language comprehension. For instance, we have created an autosuggestion model that allows users to query an insurance specific knowledge base by writing semantic questions in a search bar (product is used by Natixis). We also use NLP algorithms to detect specific entities in any piece of text, or to predict its content. We use these algorithms to categorize texts (emails or any written documents) and analyze their content. Modern NLP algorithms are Deep Learning algorithms, i.e. they are built upon multiple layers with millions or trainable parameters.
  • Document2Insights: We use the latest computer vision AI models to efficiently detect, locate and classify the documents that are sent through the API. We have then custom OCR models (both standard & handwritten) to read the documents and return structured information from the different fields extracted.
    Some neural networks are trained to recognize non-textual elements, such as checkboxes, signatures, drawings, logos, etc. Both ML based and standard image processing methods are used to preprocess the images and enhance results, or generate synthetic data to increase training datasets.
    Computer vision algorithms are also combined with NLP models to perform document classification, and entities extraction (NER).
  • ARM: we use ARM associated rule mining. In our case, this can be used to find patterns in the profiles of customers buying insurance products, whether in the profile of the person, or in the type of products they buy. It helps providing explainability to the agents about the prediction and recommendation we’re pushing
  • Library Shap: Depend of the importance of the variable value (e.g: age=the most important variable)
  • Shap Local: explain the prediction of an instance x by computing the contribution of each feature to the prediction
  • SkopeRules: GBoost with decision trees

About Zelros

Founded in 2016 by Christophe Bourguignat, Fabien Vauchelles and Damien Philippon, Zelros is a B2B independent software vendor developing Artificial Intelligence for insurance and bank insurance players. Zelros’ mission is to enable insurance players to revolutionize and re-enchant their relationship with their customers by letting them take ownership of AI. For more information, please visit

Source: Zelros

→ back to the InsurTech Spotlight index