It is hard to imagine that AstraZeneca, as an international pharmaceutical giant, once relied on a team of more than a dozen people to complete compliance consultation. In daily life, they need to complete the online consultation of 8000+ medical representatives, and deal with a lot of repeated reimbursement, meetings and customer reception problems. Among them, 80% medical representatives will ask 20% repeated questions. After consultation, medical representatives often ask compliance colleagues to make a second reply online to collect relevant texts.
At the end of the month or on the promotion day of the event, the online questions received by the department reached a peak. At that time, a dozen compliance personnel were often too busy to eat.
For enterprises, it is undoubtedly a waste of talents to set up a special person to answer questions day and night. Busy and mechanical basic work is killing the will of practitioners. At the same time, recruiting suitable people has become a difficult problem for human resources departments.
Therefore, it is an urgent problem for AstraZeneca to improve the efficiency of enterprises and replace the original service model through new technologies. The maturity of artificial intelligence technology provides the possibility for the intelligence of information consultation. It is of great strategic significance for AstraZeneca's compliance personnel to welcome new students. ...
The medical field is highly specialized, especially for the strict order department such as AstraZeneca Compliance Department, the rational use of technical terms is very important, which will directly affect the accuracy of the reply results. Therefore, how to ensure the accuracy of the input language of artificial intelligence products and ensure enough professional speech skills is the most important problem that AstraZeneca compliant robots need to overcome.
In the face of AstraZeneca's olive branch, a mysterious artificial intelligence startup accepted the contract.
In less than a month, the team asked AstraZeneca for a corpus of 70 frequently asked questions and three standard business processes. After understanding and studying the laws and regulations related to medicine, the team quickly sorted out the data and initially established a knowledge base containing more than 100 knowledge points by using AI technology.
After completing the first stage of work, the primary artificial intelligence dialogue robot has been built.
Subsequently, the robot was launched inside AstraZeneca and conducted a "clinical trial session" through relevant sales staff. This process, like a withered sponge, is put in the water. During the two-month test, new data and questions became the "nutrition fast line" of the dialogue robot.
Its core knowledge base is constantly expanding. When the whole product is "mature", the knowledge base already contains about 374 knowledge points and 5232 questions, which can cover 82.3% of the questions raised by employees and medical representatives, and the data is still rising.
This formed AI dialogue robot consists of three modules: robot question-and-answer module, form module and task cluster module. They correspond to the three dialogue modes between people respectively, and are uniformly regulated by the intelligent center console.
Specifically, the intelligent center console will classify and identify the collected questions, analyze their dialogue patterns, and assign tasks to modules suitable for handling tasks.
"Robot Q&A module" mainly deals with simple Q&A dialogue, that is, answering one-dimensional questions raised by users in a single context. For this kind of problem, the robot will accurately match the problem according to the knowledge base and solve the common problems such as meeting time query effectively.
More dialogues in the medical field are not simple one-dimensional dialogues, and often involve product description and comparison, which means that AI and users must limit the chat content to a certain range (such as a table) in order to communicate normally, which requires the use of "table module" for processing.
For example, when consulting AstraZeneca products, medical representatives often need to obtain drug information at a limited price and compare different drugs. For such questions, AstraZeneca can enter relevant price statements (such as Excel) into the system in advance to help medical representatives confirm the communication scope and complete multiple rounds of questions for specific information.
In this way, when the user communicates with the AI dialogue robot, the robot can understand what kind of problems the user refers to by combining relevant information, and then realize the query, sorting and screening of relevant information based on tables, solve the sorting and output of two-dimensional information, and solve the problems of drug price comparison and expiration date comparison.
The third module "Task Cluster Module" is used to solve specific problems. Some users may ask some specific questions, such as trying to consult the project application through the AI ? ? dialogue robot system. For this kind of problem, the AI dialogue robot will guide the user to input some information needed for the application, answer the questions in the application process and guide the user to complete the relevant procedures.
At present, the accuracy of this AI dialogue robot system has exceeded 90%, and the trigger rate of intelligent recommendation has reached 98%. More than a dozen members of AstraZeneca's compliance department now have time and energy to do more valuable work.
This mysterious AI team is called Leia. As can be seen from official website, the company was founded on 20 15, initiated by a doctoral and MBA team who returned from Ivy League, and is committed to becoming an intelligent robot company with global influence in the era of man-machine life.
Before contacting AstraZeneca's compliance robot project, Leia's customers included Ctrip, Wyeth and other large To C enterprises. However, after communicating with AstraZeneca, it is found that AstraZeneca is different from those enterprises facing C-end customers.
For the To C project, when designing the AI dialogue robot, we often put more energy into the consumer experience, making the whole chat more interesting and improving the customer conversion rate. However, AstraZeneca is an a To B company, so it pays more attention to the professionalism and accuracy of future AI.
From the results, it took only three months to build and polish the whole AI product, but more than 50% of the cost was reduced. For pharmaceutical and medical-related enterprises, it is obviously a very cost-effective thing to optimize manpower, reduce costs and realize digital management through dialogue robots.
Moreover, according to Han Rui, project director and commercial director of Leye AstraZeneca, "In the process of cooperation with AstraZeneca, we also found some other cooperation possibilities. In some remote areas, medical representatives always shirk the number of trips to the area because the distance is too far. Therefore, when local doctors encounter problems or needs, they can't contact medical representatives in time to solve the problems. For pharmaceutical companies, this will undoubtedly cause sales losses. In this regard, Lai is also planning to build an ai dialogue robot for doctors, that is, to provide professional technical support for doctors in hospitals covered by pharmaceutical companies through mobile phone ports. "
For example, when a doctor needs medicine, he can consult the robot first. On the one hand, the AI robot can answer the doctor's basic questions, on the other hand, it will inform the medical representative in charge of this area to remind him to return to the relevant area in time. This model can significantly increase the activity and stickiness of doctors on the platform of pharmaceutical companies and bring more business opportunities to pharmaceutical companies.
What's more worth mentioning is that the model of using small knowledge base to solve practical problems and commercializing it has also been developed in the past, which has gradually made AI technology move from industrialization to consumers and gradually solved the daily problems of consumers and staff.
In this regard, Han Rui said: "At present, robots have taken root in medicine, maternal and child industries. In the future, when the AI-based knowledge base industry becomes more and more abundant and more enterprises benefit, the digital era of enterprise management will really come. "