Applying AIGC to Transform Insurance Marketing at Ant Group

Introduction – With rapid advances in artificial intelligence, the insurance industry’s marketing models are evolving. Ant Group’s insurance marketing growth algorithm team uses AIGC (Artificial Intelligence‑Generated Content) to generate personalized marketing assets, automate content recommendation, and produce various video formats, enhancing both efficiency and user interaction.

1. Background – Traditional insurance marketing struggles to meet the growing demand for personalized, large‑scale content creation, product recommendation, and complex policy explanation. Manual processes cannot satisfy diverse consumer preferences, prompting the algorithm team to build an automated, intelligent, and highly personalized marketing ecosystem powered by AIGC.

Ant’s marketing assets fall into two categories: (a) product‑linked cards and short copy generated by AIGC, and (b) hotspot‑related content such as influencer news and short videos that drive traffic to Ant Insurance.

2. Overall Solution

Professional & Personalized GC Capability – Different users have different concerns (e.g., coverage breadth vs. claim limits). The system provides user‑specific product introductions, requiring both professional accuracy and personalization.

The workflow begins by defining a concrete scenario (e.g., card type or recommendation reason), selecting a template, and inputting product information and style. User tags are fetched from a pre‑generated audience pool, and the product knowledge base is consulted for the first‑level demand inference. Hotspot events are also matched when relevant (e.g., increased demand for child insurance during a flu outbreak).

Based on audience profiles, product needs, professional knowledge, and matched hotspots, the system generates the appropriate copy, which is then reviewed, edited, or refined through an interactive interface before final deployment.

Key Steps in Personalized Copy Production

Quality Evaluation – Ensuring professional accuracy through quality monitoring, compliance models, safety inspections, and human review, with full traceability.

Online Replacement – Filtering out low‑performing generated copies and continuously optimizing matching processes.

Effect Recovery – Collecting performance data to enrich the argument library for better future inference.

Hotspot content generation follows a three‑step process: (1) filter event streams for insurance‑related topics and label them using CoT + Prompt with large models; (2) generate assets (rewritten articles, long‑form graphics, videos) using RAG combined with insurance knowledge; (3) conduct strict quality and compliance checks before publishing.

3. Future Outlook – As large‑model technology matures, it will enable more inclusive insurance services, compressing insurance and medical knowledge, solving complex decision problems, and delivering rigorous, professional, and personalized interactions that improve user experience and drive product innovation.

In summary, AIGC not only boosts content generation efficiency and quality but also propels insurance product innovation, with broader and deeper applications expected in the future.