Author | Ronglian Cloud
Proofreading | Wang Yaxuan
Recently, the QCon Global Software Development Conference was officially held. Tang Xingcai, the product leader of Ronglian Cloud's large model, was invited to attend and share the innovative application and practice of large models in marketing service scenarios.
Tang Xingcai pointed out that driven by the wave of large models, the marketing service scene is undergoing unprecedented changes. In the face of fierce market competition and the diversification of consumer needs, enterprises urgently need to use large model technology to improve efficiency and reduce costs in actual business scenarios, while meeting the needs of users for personalized services.
Challenges faced by corporate marketing services
Challenge 1
With the increase in economic uncertainty, the slow recovery of the marketing environment and market, the purchasing power of consumers is affected, the cost of customer acquisition is getting higher and higher, and the ROI is getting lower and lower.
Challenge 2
Consumers' demand for products and services is becoming more and more personalized, which requires that in the marketing service chain, we should not only pay attention to the quality of products and the experience of services, but also be able to provide solutions that meet the personalized needs of consumers.
Opportunities brought about by large-scale model technology
Ronglian Cloud comprehensively considers the business value in the marketing service scenario, how to use large model technology to improve customer service experience and precision marketing, and combines the feasibility of the current large model technology and actual business scenarios to train a professional large model assistant with industry attributes.
Large-scale model practice case sharing
1. Financial customer service assistant: improve service efficiency
In the financial industry, the customer service team undertakes full-link business before, during and after the loan, with few people and heavy tasks, and the amount of knowledge and standardized content involved is huge. In order for the structure and content of knowledge to be usable, it requires extremely high labor costs, disassembly and maintenance, and the maintenance and updating of knowledge are extremely expensive. How agents can quickly and accurately retrieve relevant knowledge to provide efficient customer service. and how to control the personnel investment and cost control of AI engineers.
Rongxi Copilot has launched a large-model customer service assistant, and the intelligent knowledge base can greatly improve the efficiency of knowledge retrieval, reduce the retrieval time of complex problems of customer service personnel from minutes to seconds, and significantly improve customer service efficiency and customer satisfaction. On the O&M side, the whole process has changed from manual work to automatic QA mining of large models and put into use in seconds. Truly help enterprises reduce costs and increase efficiency.
For knowledge, it is only valuable when it flows. First of all, we can quickly disassemble massive enterprise knowledge and generate a real-time updated enterprise knowledge base, so that employees can ask questions to the customer service assistant of the Copilot model anywhere, and the knowledge accuracy rate reaches 89%.
2. Bank Assistant: Increase business conversion rate
In the banking marketing business, excellent marketing skills can effectively improve the business conversion rate. Originally, it was up to the supervisor to summarize the speech to the marketer after listening to a large number of recordings, but this method was time-consuming and laborious, and it could not exhaust and comprehensively obtain the gold medal strategy. Later, an AI algorithm was introduced to extract words, which solved the problem of quantity, but the quality of the algorithm was limited because of the lack of business experience and judgment of the effectiveness of the dialogue.
Rongxi Copilot has launched a bank phone assistant, which uses large-scale model speech mining technology to quickly extract gold medal words from massive call data, and sets a call extraction target, which can mine massive high-quality gold call words according to the conversion effect, helping agents quickly master and use flexible speech strategies, thereby significantly improving business conversion rate.
3. Group Employee Assistant: Shorten the time to solve problems
For large conglomerates with multiple subsidiaries, employees encounter a variety of administrative manpower consulting issues in their day-to-day work. However, the number of questions and answers and the scope of services supported by the bot/knowledge base are limited, which cannot meet the actual administrative manpower consulting needs of employees. The existing knowledge update and maintenance also rely heavily on manual work, and trainers need to be arranged to maintain knowledge items.
Rongxi Copilot launched the Group's employee assistant, which combines large models with RAG technology, which can accurately understand employees' intentions and quickly retrieve answers, greatly reducing problem solving time and improving employee work efficiency.
4. Intelligent work orders in the manufacturing industry: improve the accuracy of fault detection
In the manufacturing industry, the fault descriptions reported by customers are not uniform, which brings challenges to after-sales service. The customer hopes to build a set of after-sales knowledge platform, and empower their after-sales voice customer service and APP online customer service scenarios through the knowledge retrieval capabilities of the large model knowledge base. At present, there are 100,000+ product manuals and manuals on the customer side, but the document structure is irregular. These complex structures cannot be directly extracted through the system, and need to rely on manual processing to extract, but hundreds of thousands of orders of data cannot be processed manually, and it is hoped that this kind of document knowledge can be used through large-scale model capabilities.
Rongxi Copilot large model intelligent work order can accurately understand the fault description in the work order, extract equipment fault information, improve the accuracy of fault judgment, and reduce manual processing costs.
Future outlook
Although large model technology brings great potential, it also faces challenges such as high computing costs, difficult to calculate ROI, and difficult to change existing business processes. In the future, we look forward to reducing costs and improving efficiency through strategies such as localized GPU all-in-one machines, small-parameter models, and large-scale models. At the same time, it will increase the talent pool, take business goals as the guide, gradually realize the upgrade of application layer capabilities, and guide the optimization and innovation of business processes.
In this era of intelligent innovation, large-scale model technology has undoubtedly brought new vitality to the marketing service industry. Let's look forward to the fact that in the near future, the large-scale model application launched by Ronglian Cloud can better serve enterprises and promote the development of the entire industry in a more efficient and intelligent direction.
Article reprint source:
httpstps://www.infoc.cn/articlel/Obxvallm17B4VesSquare