A digital assistant system

Abstract

The present invention relates to a system (1) which increases operational efficiency by reducing the access time of dealership employees to information, minimizes human-induced errors, and automates the processes of order management, delivery estimates, technical documentation access, target information, and dealership communication of dealership employees by using a Retrieval-Augmented Generation (RAG) structure.

Description

[0001] DESCRIPTION

[0002] A DIGITAL ASSISTANT SYSTEM

[0003] Technical Field

[0004] The present invention relates to a system which increases operational efficiency by reducing the access time of dealership employees to information, minimizes human-induced errors, and automates the processes of order management, delivery estimates, technical documentation access, target information, and dealership communication of dealership employees by using a Retrieval-Augmented Generation (RAG) structure.

[0005] Background of the Invention

[0006] Today, methods such as manual processes, conventional CRM and ERP systems, classic chatbot solutions, and static knowledge bases are generally used for dealership employees to access information and manage operational processes:

[0007] • Static Knowledge Bases: Dealerships generally try to access the data they need by searching for information through in-house portals or documentation systems. However, these systems are difficult to update, time-consuming to access, and limited in terms of user experience.

[0008] • CRM and ERP Systems: Dealership employees must log into CRM or ERP systems in order to access order, delivery, and target information. However, these systems are often complex, making it difficult to quickly access the accurate information.

[0009] • Manual Call and Email Traffic: Dealership employees often have to contact call centers or wait for email responses to access technical information or order / estimated delivery details. These processes lead both to time loss and human-induced errors.• Classic Chatbot Solutions: Some chatbot solutions used work with simple keyword matching techniques and cannot provide contextually accurate responses to specific questions from users.

[0010] Current methods have certain shortcomings and disadvantages, which are:

[0011] • Time Loss and Inefficiency: Logging into different platforms and contacting call centers to access information, as well as performing manual searches in static systems, slows down processes.

[0012] • Outdated or Incomplete Information: Since conventional knowledge bases are static, they may contain outdated content, which increases the risk of making decisions based on incorrect information.

[0013] • Lack of Contextual and Natural Language Processing: Since older chatbot solutions only work with specific patterned sentences, they cannot respond flexibly and accurately enough to questions asked by users in natural language.

[0014] • Non-Customized Solutions for Dealership Organizations: General-purpose chatbots and knowledge management systems are not specifically optimized for automotive dealerships and fall short in prioritizing the specific information that dealership employees need.

[0015] • Operational Costs: The extra time spent by dealership employees to access information due to manual processes and call center costs pose an additional burden for businesses.

[0016] For this reason, there is a need for a new system which will overcome the above-mentioned shortcomings and optimize operational processes by providing the dealership organization with fast, accurate, and personalized information.

[0017] The international patent document no. WO2024226755, an application included in the state of the art, discloses a system which uses graph-based Natural Language Processing (NLP) for querying, analyzing, and visualizing complex data structures. Such a system runs a generalized Al language model; defines and migrates atraining dataset into a graph database by exposing data sources to an executing Al language model that self-identifies a structure and self-writes an executable script to query the original data sources and self-writes code to load the extracted data into a graph database in the form of new nodes and new relationships with directionality between nodes. The system further includes means for loading the extracted data into a graph database; condensing the information stored within the graph database into a condensed data structure representing the full architecture of the data in natural language format; and responding to human language queries with responsive text, speech, and visualizations using the data loaded into the graph database. Additionally, the system uses GPT, LLM, and RAG while performing data analysis to respond to queries.

[0018] However, the said system is not a digital assistant system focused on optimizing the operational processes of dealership employees. The primary object of the system is not to accelerate and facilitate important operational processes such as order management, delivery estimates, target information, and access to technical documentation for dealership employees. It does not offer real-time decision support mechanisms focused on dealership operations.

[0019] Summary of the Invention

[0020] An object of the present invention is to realize a system which increases operational efficiency by reducing the access time of dealership employees to information, minimizes human-induced errors, and automates the processes of order management, delivery estimates, technical documentation access, target information, and dealership communication of dealership employees by using a Retrieval-Augmented Generation (RAG) structure.

[0021] Another object of the present invention is to realize a system which enables saving time by automating access of dealership employees to order, delivery and target information.A further object of the present invention is to realize a system which enables contextual and accurate responses to be given to questions asked by users in natural language, up-to-date and reliable information to be provided through real-time data integration, and operational processes to be improved by minimizing human-induced errors.

[0022] Detailed Description of the Invention

[0023] “A Digital Assistant System” realized to fulfil the objectives of the present invention is shown in the figure attached, in which:

[0024] Figure l is a schematic view of the inventive system.

[0025] The components illustrated in the figure are individually numbered, where the numbers refer to the following:

[0026] 1. System

[0027] 2. Electronic device

[0028] 3. Classification server

[0029] 4. Technical documentation server

[0030] 5. Target information server

[0031] 6. Order / delivery information server

[0032] 7. External dealership communication server

[0033] 8. Daily chat server

[0034] 9. Server

[0035] The inventive system (1) which increases operational efficiency by reducing the access time of dealership employees to information, minimizes human-induced errors, and automates the processes of order management, delivery estimates, technical documentation access, target information, and dealership communicationof dealership employees by using a Retrieval-Augmented Generation (RAG) structure and artificial intelligence comprises

[0036] at least one electronic device (2) which is used by the dealership employee and enables dealership employees to ask questions and receive answers in order to obtain information regarding operational processes;

[0037] at least one classification server (3) which decides by which server the question received from the dealership employee should be answered; at least one technical documentation server (4) which enables technical questions to be answered with RAG architecture and semantic / keyword search (hybrid search);

[0038] at least one target information server (5) which generates monthly / quarterly target information of the user with REST API (Representational State Transfer Application Programming Interface) calls and GPT (Generative Pre-trained Transformer) model;

[0039] at least one order / delivery information server (6) which optimizes the status and estimated delivery date information of orders regarding purchases made from dealerships via GPT by retrieving them with REST API;

[0040] at least one external dealership communication server (7) which provides the contact information of the dealership the user wishes to reach;

[0041] at least one daily chat server (8) which manages general chat purpose questions with GPT; and

[0042] at least one server (9) which is configured to decide, when a question is received from a dealership employee, to which server this question should be forwarded through the classification server (3); to receive the response generated by the GPT model by performing semantic interpretation by using data sources in the form of REST API, databases, technical documents, and past user interactions of the selected server, and to quickly forward it to dealership employee; to generate responses improved by dealership employee interactions; and to optimize costs by using cached responses if the question asked has been answered before.The electronic device (2) included in the inventive system (1) is a device that is a smartphone, tablet computer, or portable computer allowing at least one application to be run thereon. The said electronic device (2) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to enable data exchange with the server (9) through this established connection. In the preferred embodiment of the invention, the electronic device (2) is configured to exchange data with the server (9) by using the Internet as a data bus.

[0043] The classification server (3) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The classification server (3) is configured to enable incoming questions to be forwarded to the relevant module by using the GPT model and pre-trained classifiers. In one embodiment of the invention, the classification server (3) is configured to enable precise classification to be performed by using BERT or T5-based models. In one embodiment of the invention, the classification server (3) is configured to directly forward simple queries to specific servers via rule-based NLP systems. In one embodiment of the invention, the classification server (3) is configured to enable flexible classification to be performed via zero-shot learning or few-shot learning methods.

[0044] The technical documentation server (4) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The technical documentation server (4) is configured to retrieve accurate information from technical documents through RAG (Retrieval-Augmented Generation) and semantic search / keyword search (hybrid search). In one embodiment of the invention, the technical documentation server (4) is configured to enable pure semantic search to be performed with vector databases in the form of Pinecone, FAISS, and Weaviate. Inanother embodiment of the invention, the technical documentation server (4) is configured to enable large-scale document scanning to be performed by using Google BERT-based retriever models. In a further embodiment of the invention, the technical documentation server (4) is configured to enable technical documents to be processed automatically by creating a rule-based knowledge graph in the form of a Knowledge Graph.

[0045] The target information server (5) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The target information server (5) is configured to summarize and analyze target information of dealership users via the REST API and GPT model. In one embodiment of the invention, the target information server (5) is configured to enable data to be directly retrieved via static data tables and SQL queries. In another embodiment of the invention, the target information server (5) is configured to enable information to be retrieved through a pre-created Power BI or Tableau dashboard. In a further embodiment of the invention, the target information server (5) is configured to analyze target information more flexibly by using graph-based data structures in the form of GraphQL and Neo4j .

[0046] The order / delivery information server (6) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The order / delivery information server (6) is configured to provide order status in natural language via REST API and GPT. In one embodiment of the invention, the order / delivery information server (6) is configured to retrieve information directly from the ERP via RPA (Robotic Process Automation) systems. In another embodiment of the invention, the order / delivery information server (6) is configured to enable order and delivery times to be estimated through artificial intelligence-based estimation algorithms in the form of LSTMs (Long Short-Term Memory), ARIMAs (AutoregressiveIntegrated Moving Average), and Prophet. The order / delivery information server (6) is configured to enable order information to be presented directly to the user by providing dealerships with a self-service chatbot interface.

[0047] The external dealership communication server (7) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The external dealership communication server (7) is configured to retrieve dealership information via REST API and to generate natural language responses with the GPT model. In one embodiment of the invention, the external dealership communication server (7) is configured to enable direct communication information to be provided through integration with static dealership directories in the form of Google My Business and SAP CRM. In another embodiment of the invention, the external dealership communication server (7) is configured to enable dealership contact information to be provided via voice responses by adding voice assistant integration. In a further embodiment of the invention, the external dealership communication server (7) is configured to enable additional information regarding the dealership in the form of location and stock status to be presented by using contextual recommendation engines in the form of graph-based recommendation systems.

[0048] The daily chat server (8) included in the inventive system (1) is configured to establish a connection with the server (9) by using any remote communication protocol included in the state of the art and to exchange data with the server (9) through this established connection. The daily chat server (8) is configured to manage general conversations via the GPT model. In one embodiment of the invention, the daily chat server (8) is configured to use rule-based chatbot systems in the form of Dialogflow, Rasa, and Microsoft Bot Framework. In another embodiment of the invention, the daily chat server (8) is configured to enable low-cost chat to be established via a specially trained small language model in the form of LLM distillation, TinyBERT, or T5 Small. In a further embodiment of theinvention, the daily chat server (8) is configured to enable instant responses to be generated via small artificial intelligence models running on the device in the form of Edge Al.

[0049] The server (9) included in the inventive system (1) is configured to exchange data with the classification server (3), the technical documentation server (4), the target information server (5), the order / delivery information server (6), the external dealership communication server (7), and the daily chat server (8). The server (9) is configured to enable queries made through the electronic device (2) to be received, and to be forwarded to the classification server (3) in order to determine which module the query should be forwarded to in order to determine which server will respond to the received queries; queries to be correctly forwarded via the RAG structure; responses returned to queries to be received and transmitted to the dealership employee through the electronic device (2); responses to be developed and improved by learning in line with user interactions; the accuracy rate of responses to be increased by performing semantic interpretation; and the access time of dealership employees to information to be reduced.

[0050] Industrial Application of the Invention

[0051] The most significant advancement achieved with the inventive system (1) is the creation of a digital assistant that provides customized and real-time information based on user needs, with an artificial intelligence agent (Al Agent) approach having a Retrieval-Augmented Generation (RAG)-based hierarchical module structure. Instead of the user accessing different platforms for different information, through the developed Al Agent structure, different platforms are accessed with the help of artificial intelligence on behalf of the user, and the necessary information is obtained quickly and accurately.

[0052] The inventive system (1) develops a multi-layered decision-making and response generation mechanism. It automatically analyzes the type of problem based on theinput from the user and activates the correct module. This increases the accuracy of responses while preventing the use of unnecessary system resources. By using the RAG approach in technical documentation and information inquiry processes, the powers of configured data sources and generative artificial intelligence are combined. In this way, responses are generated quickly, accurately, and in a contextually appropriate manner. The system (1) has customized integrations in order to generate real-time responses with information from in-house and external databases. Operational data such as target, order, and delivery information is automatically retrieved from in-house systems through REST API integrations and dynamically presented to the user. The order / delivery information server (6) provides more accurate and predictable delivery estimates by leveraging “Digital Warehouse” data developed in logistics processes. The system (1) generates instant and context-appropriate responses by accurately analyzing user needs. It provides a personalized experience for each user by accurately delivering personalized information such as dealership target data. The system (1) revolutionizes operational processes by automating manual processes. By reducing information access times by more than 50%, it lightens the operational load on users. By means of automation, error rates arising from manual processes are minimized.

[0053] Within these basic concepts; it is possible to develop various embodiments of the inventive “A Digital Assistant System (1)”; the invention cannot be limited to examples disclosed herein and it is essentially according to claims.

Claims

CLAIMS1. A system (1) which increases operational efficiency by reducing the access time of dealership employees to information, minimizes human-induced errors, and automates the processes of order management, delivery estimates, technical documentation access, target information, and dealership communication of dealership employees by using a Retrieval-Augmented Generation (RAG) structure and artificial intelligence; characterized byat least one electronic device (2) which is used by the dealership employee and enables dealership employees to ask questions and receive answers in order to obtain information regarding operational processes;at least one classification server (3) which decides by which server the question received from the dealership employee should be answered; at least one technical documentation server (4) which enables technical questions to be answered with RAG architecture and semantic / keyword search (hybrid search);at least one target information server (5) which generates monthly / quarterly target information of the user with REST API (Representational State Transfer Application Programming Interface) calls and GPT (Generative Pre-trained Transformer) model;at least one order / delivery information server (6) which optimizes the status and estimated delivery date information of orders regarding purchases made from dealerships via GPT by retrieving them with REST API;at least one external dealership communication server (7) which provides the contact information of the dealership the user wishes to reach;at least one daily chat server (8) which manages general chat purpose questions with GPT; andat least one server (9) which is configured to decide, when a question is received from a dealership employee, to which server this question should be forwarded through the classification server (3); to receive the response generated by the GPT model by performing semantic interpretation by usingdata sources in the form of REST API, databases, technical documents, and past user interactions of the selected server, and to quickly forward it to dealership employee; to generate responses improved by dealership employee interactions; and to optimize costs by using cached responses if the question asked has been answered before.

2. A system (1) according to Claim 1; characterized by the electronic device (2) which is a device that is a smartphone, tablet computer, or portable computer allowing at least one application to be run thereon.

3. A system (1) according to Claim 1 or 2; characterized by the electronic device (2) which is configured to establish a connection with the server (9) by using any remote communication protocol and to enable data exchange with the server (9) through this established connection.

4. A system (1) according Claim 3; characterized by the electronic device (2) which is configured to exchange data with the server (9) by using the Internet as a data bus.

5. A system (1) according to any one of the preceding claims; characterized by the classification server (3) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

6. A system (1) according to any one of the preceding claims; characterized by the classification server (3) which is configured to enable incoming questions to be forwarded to the relevant module by using the GPT model and pre-trained classifiers.

7. A system (1) according to any one of the preceding claims; characterized by the classification server (3) which is configured to enable precise classification to be performed by using BERT or T5-based models.

8. A system (1) according to any one of the preceding claims; characterized by the classification server (3) which is configured to directly forward simple queries to specific servers via rule-based NLP systems.

9. A system (1) according to any one of the preceding claims; characterized by the classification server (3) which is configured to enable flexible classification to be performed via zero-shot learning or few-shot learning methods.

10. A system (1) according to any one of the preceding claims; characterized by the technical documentation server (4) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

11. A system (1) according to any one of the preceding claims; characterized by the technical documentation server (4) which is configured to retrieve accurate information from technical documents through RAG (Retrieval-Augmented Generation) and semantic search / keyword search (hybrid search).

12. A system (1) according to any one of the preceding claims; characterized by the technical documentation server (4) which is configured to enable pure semantic search to be performed with vector databases in the form of Pinecone, FAISS, and Weaviate.

13. A system (1) according to any one of the preceding claims; characterized by the technical documentation server (4) which is configured to enable large-scale document scanning to be performed by using Google BERT-based retriever models.

14. A system (1) according to any one of the preceding claims; characterized by the technical documentation server (4) which is configured to enable technical documents to be processed automatically by creating a rule-based knowledge graph in the form of a Knowledge Graph.

15. A system (1) according to any one of the preceding claims; characterized by the target information server (5) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

16. A system (1) according to any one of the preceding claims; characterized by the target information server (5) which is configured to summarize and analyze target information of dealership users via the REST API and GPT model.

17. A system (1) according to any one of the preceding claims; characterized by the target information server (5) which is configured to enable data to be directly retrieved via static data tables and SQL queries.

18. A system (1) according to any one of the preceding claims; characterized by the target information server (5) which is configured to enable information to be retrieved through a pre-created Power BI or Tableau dashboard.

19. A system (1) according to any one of the preceding claims; characterized by the target information server (5) which is configured to analyze target information more flexibly by using graph-based data structures in the form of GraphQL and Neo4j .

20. A system (1) according to any one of the preceding claims; characterized by the order / delivery information server (6) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

21. A system (1) according to any one of the preceding claims; characterized by the order / delivery information server (6) which is configured to provide order status in natural language via REST API and GPT.

22. A system (1) according to any one of the preceding claims; characterized by the order / delivery information server (6) which is configured to retrieve information directly from the ERP via RPA (Robotic Process Automation) systems.

23. A system (1) according to any one of the preceding claims; characterized by the order / delivery information server (6) which is configured to enable order and delivery times to be estimated through artificial intelligence-based estimation algorithms in the form of LSTMs (Long Short-Term Memory), ARIMAs (Autoregressive Integrated Moving Average), and Prophet.

24. A system (1) according to any one of the preceding claims; characterized by the order / delivery information server (6) which is configured to enable order information to be presented directly to the user by providing dealerships with a self-service chatbot interface.

25. A system (1) according to any one of the preceding claims; characterized by the external dealership communication server (7) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

26. A system (1) according to any one of the preceding claims; characterized by the external dealership communication server (7) which is configured to retrieve dealership information via REST API and to generate natural language responses with the GPT model.

27. A system (1) according to any one of the preceding claims; characterized by the external dealership communication server (7) which is configured to enable direct communication information to be provided through integration with static dealership directories in the form of Google My Business and SAP CRM.

28. A system (1) according to any one of the preceding claims; characterized by the external dealership communication server (7) which is configured to enable dealership contact information to be provided via voice responses by adding voice assistant integration.

29. A system (1) according to any one of the preceding claims; characterized by the external dealership communication server (7) which is configured to enable additional information regarding the dealership in the form of location and stock status to be presented by using contextual recommendation engines in the form of graph-based recommendation systems.

30. A system (1) according to any one of the preceding claims; characterized by the daily chat server (8) which is configured to establish a connection with the server (9) by using any remote communication protocol and to exchange data with the server (9) through this established connection.

31. A system (1) according to any one of the preceding claims; characterized by the daily chat server (8) which is configured to manage general conversations via the GPT model.

32. A system (1) according to any one of the preceding claims; characterized by the daily chat server (8) which is configured to use rule-based chatbot systems in the form of Dialogflow, Rasa, and Microsoft Bot Framework.

33. A system (1) according to any one of the preceding claims; characterized by the daily chat server (8) which is configured to enable low-cost chat to beestablished via a specially trained small language model in the form of LLM distillation, TinyBERT, or T5 Small.

34. A system (1) according to any one of the preceding claims; characterized by the daily chat server (8) which is configured to enable instant responses to be generated via small artificial intelligence models running on the device in the form of Edge Al.

35. A system (1) according to any one of the preceding claims; characterized by the server (9) which is configured to exchange data with the classification server (3), the technical documentation server (4), the target information server (5), the order / delivery information server (6), the external dealership communication server (7), and the daily chat server (8).

36. A system (1) according to any one of the preceding claims; characterized by the server (9) which is configured to enable queries made through the electronic device (2) to be received, and to be forwarded to the classification server (3) in order to determine which module the query should be forwarded to in order to determine which server will respond to the received queries; queries to be correctly forwarded via the RAG structure; responses returned to queries to be received and transmitted to the dealership employee through the electronic device (2); responses to be developed and improved by learning in line with user interactions; the accuracy rate of responses to be increased by performing semantic interpretation; and the access time of dealership employees to information to be reduced.

Images