Customers have already moved online, yet bancassurance's operating system remains anchored offline. Lu Min's solution is AI.
On September 22, 2026, the 7th Bancassurance Cooperation and Development Forum and Bancassurance Awards Ceremony was held in Beijing, hosted by Yiqu Finance Media and Insurance Enterprise Senior Advisor, under the theme "The Posture of Attack and Defense, Harmony in Light and Dust." At the event, regulatory leaders and veteran figures from banking and insurance discussed the development of bancassurance cooperation following "Document No. 65." Lu Min, former Chief Insurance Business Executive of Ping An Group, former Chairman of Autohome, and founder of Dongbaohui, judged that the industry adjustment resembles a generational shift and proposed rebuilding the entire chain of bancassurance customer operations with AI.
Lu Min judged that the difficulties facing the industry may not stem from an ordinary cycle but may be a generational replacement. Customer consumption habits have migrated online, while bancassurance's customer operations and management systems are still built on an offline foundation, with the chain broken at multiple points. Lu Min pointed out that banks and insurance companies must walk on two legs—supporting current business while using AI to rebuild the entire chain of customer operations, connecting private domain operations, content customer acquisition, robot invitations, intelligent conversion, practice training, and business management into one closed loop.
Industry adjustment may be a generational replacement: bancassurance must walk on two legs
The contraction of the individual agent channel has lasted for years. The life insurance individual agent workforce has dropped from approximately 9 million to around 2 million, with new business value declining in tandem. During this period, several rebounds occurred through discontinued-product hype, but the strength was limited. In July and August of this year, the bancassurance channel felt the strain, and the individual agent channel faced the same situation. The industry talks of "riding out the cycle," meaning maintaining existing actions and waiting for the cycle to return so business naturally recovers.
Lu Min does not agree. The industry track has already changed. The current adjustment is more like the moment over a century ago when horse carriages were replaced by automobiles—guarding the carriage and waiting for the cycle to return will not produce results. The basis for this judgment comes from changes in where customers are. Lu Min pointed out that insurance, like finance, faces mobile internet as its biggest variable. Customers have migrated online, consumption habits are completed online, yet the operating systems of insurance companies and banks were still built over thirty years ago with an offline foundation. Although improvements have been made since, the foundation has not been replaced. With economic cycles and low interest rates overlapping, whether the original model can hold up has become a real problem.
Walking on two legs is Lu Min's recommendation. One leg first supports current business; the other leg embraces the trends of the times, using AI transformation to lead a new development path. For bancassurance's AI exploration, Lu Min starts from the most concrete difficulty.
Bancassurance's biggest difficulty is broken customer operations: private domains must follow the customer journey
Banks hold the largest customer resources. There are approximately 220,000 bank branches nationwide. The product structure is also changing. In the first half of 2026, participating insurance original insurance premium income reached 1,012.6 billion yuan, a year-on-year increase of 94.4%, with mainstream products shifting toward participating insurance. In Lu Min's view, bancassurance's difficulties center on invitation being hard and conversion being low, with the most important being broken operations. What is missing is a complete set of customer operation concepts and measure systems. The industry has long been sales-oriented, ending once the sale is complete.
Lu Min proposed that customers need to complete full life-cycle operations within private domains. Repeat purchases, referrals, and multi-product allocation all occur during continuous operations, and the customer operation chain cannot be broken. The inertia of insurance companies ending things after selling a policy also exists in banks. Some banks have begun using WeCom to invite customers into private domains. From Lu Min's observation, the supporting operating system has not been established. Most push actions ask customers to buy products and do not align with the customer journey.
Following the customer journey is a principle Lu Min repeatedly emphasized. The bancassurance funnel leaks at multiple points. Banks have enormous customer volume, the largest block of customer resources, yet utilization is very low. Invitation response is not high, and conversion after response is also not ideal. Products customers bought three years ago mature this year, with no contact in between and no connection at maturity. Low resource utilization is related to the offline operating model and needs to be supplemented by online operations. Self-centered hard selling, Lu Min judged, is likely to suffer setbacks. Customer operations must advance along the customer journey.
Five levels around the customer journey, each matched with AI tools: the five-level AI system forms a closed loop and management blind spots open layer by layer
AI implementation in the insurance industry is accelerating. According to industry exchange data, the large model adoption rate of insurance institutions rose from 26.8% in early 2025 to 44.8% in early 2026. Over these years of working on AI, Lu Min has one realization. Young people are used to starting from a single point. Lu Min's approach is to start from the system, laying out along the value chain. Only when data from each link is connected can the power of data be brought out.
Depending on the link, AI tools can be divided into five levels. The customer acquisition link uses "drones." In Lu Min's phrasing, "drone" is a figurative term pointing to a content intelligence factory offering thousand-person, thousand-face personalization. Different customers receive different content, first awakened and warmed up, then entering the invitation link. Cold invitations have very poor results and poor customer experience. Customer follow-up is what salespeople dread most; AI will prompt what each customer's next step should be. In the invitation link, different content can be mounted under the "drone," including robots, bank or company activities, small-ticket products, service tools such as policy reviews, and each salesperson's digital avatar. Robots handle insurance question answering and product comparison, with customers doing their own research and analysis. The conversion link covers needs diagnosis, plan formulation, product comparison explanations, and signing assistance. The system generates a plan before the meeting and summarizes communication progress after the meeting.
At the speech, Lu Min introduced an insurance robot that can explain products from more than 90 insurance companies. When a customer asks about a difference between two products on a certain detail, the robot can answer, while a salesperson trained only on their own company's products cannot. As bancassurance takes on more complex long-term protection products in the future, the role of robots will become even greater. In-store and not-in-store are two parallel operating lines. In Lu Min's view, whether online or offline, if the customer does not come to the store, conversion is very difficult to complete. Customers walking into offline branches are getting older, while younger customers are more accepting of online store visits, where displays and data during the process can be recorded and later used for analysis. Customers who do not come to the store are handed to digital avatars, with functions consistent with robots, supported by knowledge bases and product comparison capabilities. Post-85 customers are more willing to research on their own and then take a step forward after developing purchase intent.
The practice platform is connected with training, with practice data flowing back into training. Whichever link is weak gets strengthened, making training thousand-person, thousand-face. Lu Min admitted that a single script responding to all customers is hard to make work in a transparent information environment. Thousand-person, thousand-face matching places ever higher demands on individual sales ability, and only with AI is there a chance to achieve it.
The top level is business management. Lu Min emphasized that operating details at all levels—headquarters, branches, head office, provincial branches, and sub-branches—should be grasped. Which level has a problem can be directly located, no longer relying on layer-by-layer reporting materials. Branch operating conditions, financial manager activity levels, and behavior models are aggregated onto the platform, and improvement plans are automatically generated by the system. The supporting compliance detection is divided into six layers, covering red lines, advertising law, regulatory requirements, and more, with the finest layer checking specific products. Lu Min emphasized that whether detection can reach the product is the biggest test. The industry generally achieves 95%, Dongbaohui achieves 99.99%, and the actual requirement is 100%—saying it wrong carries suspicion of misleading.
In Lu Min's view, the five levels of tools must be used together, so that customer acquisition volume, conversion rate, team skills, and management level rise simultaneously, and business growth follows. Lu Min reminded that business in July, August, and September is expected to be poor. Whether next year can improve and whether the opening red season can be strong will also not be easy. Lu Min recalled a case from a visit to Europe more than 20 years ago: 70% of insurance in Spain is sold through banks, products cover various types of protection, the reputation is not bad, misleading behavior is less than in individual distribution, and basic risk protection for ordinary families is thus secured. Lu Min believes that all kinds of protection products being delivered to families in need through banks is what insurance should originally look like. Once AI is put to use, the prospects for bancassurance are very large.
Comments