DE102024000277A1 - Contract management device and method for training artificial intelligence - Google Patents

Abstract

The invention relates to a contract management device (1) comprising a user database (2), a legal database (3), a template database (4), and a contract generation module (5). The contract generation module (5) is configured to generate a template (6) for a contract, taking into account requirements stored in the legal database (3), and to store this template in the template database (4), and to fill out a template (6) stored in the template database (4) with user data stored in the user database (2) to generate a computer-readable contract document. The contract management device according to the invention is characterized in that the contract generation module (5) is further configured to:
- to identify changes in the user database (2) and/or the legal database (3); and
- to adapt a template (6) stored in the template database (4) taking into account changed requirements and to fill it with user data and/or changed user data to generate a revised contract document (7) and/or to fill a template (6) stored in the template database (4) with changed user data to generate a revised contract document (7) taking into account changed user data.

Description

The invention relates to a contract management device according to the preamble of claim 1 and to a method for training a first artificial intelligence that can be used by such a contract management device.

Large and small companies alike are bound to comply with legal frameworks in the course of their business activities. If applicable law and/or the information relevant to a specific contract changes, it may be necessary to draft a new contract. This task is typically entrusted to individuals with legal training, such as lawyers. Monitoring all relevant legal requirements and the data relevant to each contract involves a significant amount of manual effort. Especially if a corporation operates in different countries, different legal regulations must be adhered to.

If applicable law changes, existing contracts may need to be adjusted quickly. Furthermore, the manual synchronization of contracts with the required data represents a source of error due to human error. Information can be transferred incorrectly or even forgotten. Due to the increasing complexity of the relevant legal framework, which is reflected in the wording of the corresponding contract texts, the number of potential sources of error is increasing.

In order to reduce manual workload and corresponding errors, it is therefore desirable to provide procedures and means with which the contract creation and monitoring process can be reliably and securely automated.

A system for the automatic generation of contracts is, for example, US 2022/0391575 A1 known. The system is able to collect information relevant to a contract, supplement it if necessary, read a document template from a template database and enter the corresponding information into the template. A contract created in this way is then sent to the required parties for signature. A party entrusting the system with generating a contract can review the contract before finalization and make changes as needed. The data relevant for completing a contract can be analyzed by artificial intelligence before being transferred to the corresponding contract. If an existing contract expires, the system can plan measures to extend the contract at an early stage. The system is thus able to counteract the unwanted expiration of contracts, but cannot prevent a contract from becoming invalid due to changing external conditions.

The present invention is therefore based on the object of providing an improved contract management device.

According to the invention, this object is achieved by a contract management device having the features of claim 1. Advantageous embodiments and further developments as well as a method for training an artificial intelligence that can be used by such a contract management device emerge from the dependent claims.

• - Änderungen in der Nutzdatenbank und/oder der Rechtsdatenbank festzustellen; und
• - unter Berücksichtigung geänderter Anforderungen eine in der Vorlagendatenbank gespeicherte Vorlage anzupassen und diese mit Nutzdaten und/oder geänderten Nutzdaten zur Generierung eines überarbeiteten Vertragsdokuments auszufüllen und/oder unter Berücksichtigung geänderter Nutzdaten eine in der Vorlagendatenbank gespeicherte Vorlage mit geänderten Nutzdaten zur Generierung eines überarbeiteten Vertragsdokuments auszufüllen.

A generic contract management device, comprising a user database, a legal database, a template database and a contract generation module, wherein the contract generation module is configured to generate a template for a contract taking into account requirements held in the legal database and to store this template in the template database and to fill out a template stored in the template database with user data held in the user database to generate a computer-readable contract document, is further developed according to the invention in that the contract generation module is further configured to:

• - to detect changes in the user database and/or the legal database; and
• - to adapt a template stored in the template database taking into account changed requirements and to fill it with user data and/or changed user data to generate a revised contract document and/or to fill a template stored in the template database with changed user data to generate a revised contract document taking into account changed user data.

The contract management device according to the invention is thus not only capable of generating new contracts, but also of adapting existing contracts while monitoring applicable law and the user data relevant to the respective contracts. This prevents a contract from becoming invalid. Since this task is now performed by a corresponding computer system, Because the system is automated, the manual workload for the relevant lawyers is reduced or even eliminated entirely. Due to the high level of automation, errors caused by human error can be prevented. The contract management system is capable of processing large amounts of data in a short time and thus, if necessary, generating revised contract documents promptly. By connecting the legal database to an external source for obtaining the latest applicable requirements, the risk of overlooking legal changes can be reduced or completely eliminated.

As already described, the contract management device is a computer system. All common embodiments of computing units are possible, such as desktop computers, servers or server networks, or even mobile devices such as a smartphone, tablet computer, laptop, or the like. The user database, the legal database, and the template database are, in particular, corresponding storage areas on a computer-readable storage medium. Each can, for example, be an SQL database. However, other database systems are also possible. The contract generation module is to be understood as a purely logical module and is implemented using dedicated hardware and/or software components. For example, the contract generation module can be a computer program product comprising computer-interpretable instructions that cause the contract management device or an execution unit included in the contract management device, such as a processor, to generate corresponding revised contract documents by accessing the relevant information. The user database, legal database, and/or template database do not have to be installed in a physical unit with a corresponding processor or contract generation module, but can also be implemented externally. The corresponding databases can, for example, be stored in a data storage system comprised of a first server, while the contract generation module is executed by a second server. The corresponding servers are communicatively connected, for example, via a network such as a WAN, in particular a LAN or WLAN.

The legal database stores relevant requirements. "Requirements" refers, in particular, to the legal framework. In addition to the currently applicable requirements, outdated requirements can also be stored for future reference. Current requirements can be obtained from a suitable source via an appropriate application programming interface (API). This can be fully automated. It is particularly advantageous to maintain different requirements applicable to different domains in the legal database. "Domain" can include, for example, different countries, different legal systems, different legal areas, different product categories, and the like.

The template database stores templates generated by the contract generation module. Such a template is a computer-readable file in a specific target format. In particular, it is a text document. This text document can be in all common file formats. Each template includes, in particular, free fields that are to be filled in using the corresponding payload data stored in the payload database. The payload data includes all information relevant to a particular contract, such as the name and address of a customer; technical information about a machine relevant to the contract, e.g., a vehicle model, an engine variant, a displacement, a power output, or the like; financial information such as a leasing period or leasing rate; billing information, an IBAN, and the like.

The contract generation module is capable of monitoring the user database and the legal database and identifying corresponding changes to the user data or requirements. These are then taken into account by the contract generation module to adapt existing contracts. A correspondingly generated revised contract document can preferably be timestamped. Any changes to the contract documents can preferably be saved, allowing seamless tracking of changes.

An advantageous development of the contract management device according to the invention provides that the contract generation module is configured to use first artificial intelligence to process the content of the user database and/or the legal database and to generate the revised contract document. Artificial intelligence is reliably capable of recognizing patterns in comprehensive data sets. This increases the reliability with which relevant changes in the user data and/or the requirements are discovered. Artificial intelligence is also capable of read, here in the form of the correspondingly modified payload and/or requirements, and process this to generate new content. Thus, the first artificial intelligence is capable of creating a new template for a contract and filling it out with the respective payload to generate a revised contract document. The first artificial intelligence is pre-trained to be able to perform this task adequately. This will be discussed in more detail later. All common algorithms or models from the field of artificial intelligence are suitable.

The first artificial intelligence is preferably a pre-trained, generative transformer. In particular, it can be a large language model, also known as a large language model (LLM). Generative transformers, or large language models, are particularly reliable at processing large amounts of data, identifying relevant content, and generating new content from it. Through special pre-training in the field of legal texts, the generative transformer can also be imprinted with legal language. This ensures competent, technically accurate, and reliable formulation of the relevant legal texts—i.e., the revised contract documents to be generated.

A further advantageous embodiment of the contract management device according to the invention further provides that it has an evaluation module, wherein the evaluation module is configured to read in a revised contract document generated by the contract generation module and to mark it with an approval mark or a rejection mark. In general, particularly when using artificial intelligence, there is a risk that a revised contract document generated by the contract generation module may contain errors. It is possible to detect these errors through a manual inspection of correspondingly generated revised contract documents, for example by a lawyer. It would also be conceivable to transmit a corresponding revised contract to the respective contracting parties, who have the option of editing a revised contract document before signing it. Preferably, however, such a revised contract document is marked with an approval mark or a rejection mark using the evaluation module.

This allows the evaluation of corresponding contract documents to be further automated. The evaluation module marks a revised contract document with an approval flag if it meets all relevant requirements and is therefore suitable for signature by the contracting parties. If, however, the revised contract document contains deficiencies, for example, because user data is missing, incorrect user data is entered, the requirements are not adequately met, or due to linguistic or spelling errors, the revised contract document is marked with a rejection flag. A revised contract document marked with a rejection flag can be adjusted through a further iteration loop using the contract generation module. To do this, the contract generation module can read the rejection flag in a revised contract document imported as input data, discard the contract document and generate a new one, or directly modify the existing revised contract document.

The evaluation module, similar to the contract generation module, is a purely logical unit. This can include dedicated hardware and/or software components.

Preferably, the evaluation module is configured to use a second artificial intelligence to process the revised contract document, wherein the second artificial intelligence is trained to mark revised contract documents with an approval or rejection flag depending on their content. Thus, artificial intelligence can also be used to evaluate the revised contract documents generated by the contract generation module. In this way, corresponding contract documents can be reliably marked with an appropriate approval or rejection flag. This can further reduce the risk of an incorrect evaluation due to human error. All common algorithms and methods from the field of artificial intelligence are suitable here, in particular the use of machine learning models. A corresponding machine learning model can preferably be based on an artificial neural network.

A further advantageous embodiment of the contract management device further provides that the contract generation module is configured to read in a revised contract document marked by the evaluation module and to further train the first artificial intelligence, taking into account the approval mark or rejection mark. The approval mark is an indication to the first artificial intelligence that a corresponding revised contract document is correct. was generated correctly. The rejection flag, on the other hand, indicates that errors were made and regeneration is necessary. Taking into account the respective approval flags and rejection flags, the ability of the first artificial intelligence to generate corresponding revised contract documents is improved.

According to a further advantageous embodiment of the contract management device, the contract generation module is further configured to transmit a revised contract document, in particular a revised contract document marked with an approval mark by an evaluation module, to a user. As already mentioned, this makes it possible to manually check the correctness of corresponding revised contract documents and to edit them as needed. Furthermore, the relevant party or user can directly sign a correct revised contract document. Various options are possible for this, such as printing the contract document, signing it manually, and digitizing it again, for example by scanning. An electronic contract document can also be digitally signed, for example with a facsimile signature, a cryptographic signature, or the like.

A further advantageous embodiment of the contract management device further provides that the contract generation module is configured to save a revised contract document in the legal database and interpret this as amended requirements. This enables the contract generation module or the contract management device to interpret changes to existing contracts as changes to the existing law. This can affect other existing contracts. This makes it possible to identify situations in which changing a first contract requires changing a second contract. Accordingly, the second, third, fourth, or subsequent contracts can also be revised accordingly. This further increases the degree of automation and the reliability that all relevant contracts are correctly executed and adhered to.

• - Generieren des überarbeiteten Vertragsdokuments mittels erster künstlicher Intelligenz;
• - Verarbeiten des überarbeiteten Vertragsdokuments mittels zweiter künstlicher Intelligenz, wobei der zweiten künstlichen Intelligenz in einem Initialtraining als Grundwahrheit eine Vielzahl von manuell mit einer Freigabemarkierung oder einer Ablehnungsmarkierung versehene Trainingsvertragsdokumente zugeführt wurden;
• - Versehen des überarbeiteten Vertragsdokuments mit einer Freigabemarkierung oder einer Ablehnungsmarkierung durch die zweite künstliche Intelligenz; und
• - Zuführen des markierten überarbeiteten Vertragsdokuments der ersten künstlichen Intelligenz als Grundwahrheit.

A method for training a first artificial intelligence that can be used in the contract management device described above comprises the following method steps:

• - Generating the revised contract document using first-class artificial intelligence;
• - Processing the revised contract document by means of a second artificial intelligence, wherein a plurality of training contract documents manually marked with an approval mark or a rejection mark were fed to the second artificial intelligence as ground truth in an initial training;
• - Marking the revised contract document with an approval or rejection mark by the second artificial intelligence; and
• - Feeding the marked revised contract document to the first artificial intelligence as ground truth.

The method according to the invention thus allows the first artificial intelligence to be trained using the second artificial intelligence. The second artificial intelligence itself was trained through a supervised learning process. During initial training, the second artificial intelligence receives as input data corresponding training contract documents that have been manually marked, in particular by a lawyer, with an approval or rejection mark. This gradually teaches the second artificial intelligence to evaluate contract documents in the same way a professional lawyer would. Accordingly, after the initial training, the second artificial intelligence is suitable for evaluating the revised contract documents generated by the first artificial intelligence. The contract documents marked by the second artificial intelligence are then passed back to the first artificial intelligence, whereupon the first artificial intelligence, as already described, gradually learns to generate revised contract documents in such a way that they contain fewer and fewer errors. This is a fully automated process, which further minimizes the corresponding manual user effort.

Further advantageous embodiments of the contract management device according to the invention and of the method according to the invention for training artificial intelligence also emerge from the exemplary embodiment, which is described in more detail below with reference to the figure.

This shows 1 a flowchart of a scheme for generating revised contract documents performed by a contract management device according to the invention.

A contract management device 1 according to the invention comprises a user database 2, a legal database 3, a template database 4, a contract generation module 5, and preferably an evaluation module 8. The user database 2 stores user data relevant for generating contracts. The legal database 3 stores requirements that correspond to applicable law, particularly domain-specific ones. The template database 4 stores templates 6 for corresponding contracts, which are to be filled out with the respective user data.

Respective user data or requirements may change during the term of a contract. This may make it necessary to amend an existing contract to ensure its continued validity. The user database 2 and/or the legal database 3 can be connected to an external source 13, for example, via the Internet, to automatically retrieve corresponding changes.

The contract generation module 5 is configured to detect any changes and, taking these changes into account, adapt a template 6 stored in the template database 4 and fill it with user data and/or modified user data to generate a revised contract document 7. It is also possible for the contract generation module 5 to fill a stored template 6 with modified user data, taking modified user data into account. An adapted template 6 generated by the contract generation module 5 can also be stored in the template database 4.

Thus, existing contracts can be kept up to date easily, reliably and without manual effort using the contract management device 1, so that legal violations are avoided.

Particularly preferably, the contract generation module 5 uses first artificial intelligence AI1 to process the content of the user database 2 and/or the legal database 3 and to generate the revised contract documents 7. Particularly preferably, this is a pre-trained, generative transformer, particularly used in conjunction with a large language model.

To improve the generation quality of corresponding revised contract documents 7, the first artificial intelligence AI1 is continuously trained during its usage phase. This is achieved by the aforementioned evaluation module 8, which particularly preferably uses the second artificial intelligence AI2 to evaluate the respective revised contract documents 7. The second artificial intelligence AI2 is trained accordingly to evaluate the revised contract documents 7. During an initial training phase, training documents 12, marked with an approval mark 9 and a rejection mark 10, are manually fed to the second artificial intelligence AI2. This gradually enables the second artificial intelligence AI2 to evaluate the corresponding revised contract documents 7 like a professional lawyer.

A correspondingly marked revised contract document 7* can then be forwarded to the respective user 11 for signature and/or editing. It would also be possible to transmit an unmarked revised contract document 7 to the respective user 11.

Such a revised contract document 7* marked with an approval mark 9 or a rejection mark 10 is then sent back to the first artificial intelligence AI1 for training. Taking into account the respective approval marks 9 and rejection marks 10, the first artificial intelligence AI1 gradually learns to improve its generation quality for generating corresponding revised contract documents 7.

With the help of the contract management device 1 according to the invention and said method for training the first artificial intelligence AI1, it is possible to comply with legal framework conditions quickly, easily and reliably at all times.

QUOTES CONTAINED IN THE DESCRIPTION

This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• US 2022/0391575 A1 [0005]

Claims (9)

Contract management device (1) comprising a user database (2), a legal database (3), a template database (4) and a contract generation module (5), wherein the contract generation module (5) is configured to generate a template (6) for a contract taking into account requirements held in the legal database (3) and to store this template in the template database (4) and to fill out a template (6) stored in the template database (4) with user data held in the user database (2) to generate a computer-readable contract document, characterized in that the contract generation module (5) is further configured to: - determine changes in the user database (2) and/or the legal database (3); and - taking into account changed requirements, to adapt a template (6) stored in the template database (4) and to fill it with user data and/or changed user data to generate a revised contract document (7) and/or taking into account changed user data, to fill a template (6) stored in the template database (4) with changed user data to generate a revised contract document (7). Contract management device (1) according to Claim 1 , characterized in that the contract generation module (5) is designed to use first artificial intelligence (KI1) to process the content of the user database (2) and/or the legal database (3) and to generate the revised contract document (7). Contract management device (1) according to Claim 2 , characterized by the first artificial intelligence (AI1) in the form of a pre-trained, generative transformer. Contract management device (1) according to one of the Claims 1 until 3 , characterized by an evaluation module (8), wherein the evaluation module (8) is configured to read in a revised contract document (7) generated by the contract generation module (5) and to provide it with an approval mark (9) or a rejection mark (10). Contract management device (1) according to Claim 4 , characterized in that the evaluation module (8) is designed to use second artificial intelligence (KI2) to process the revised contract document (7), wherein the second artificial intelligence (KI2) is trained to provide revised contract documents (7) with the approval mark (9) or the rejection mark (10) depending on their content. Contract management device (1) according to Claim 2 or 3 as well as Claim 4 or 5 , characterized in that the contract generation module (5) is further configured to read in a revised contract document (7*) marked by the evaluation module (8) and to further train the first artificial intelligence (KI1) taking into account the approval marking (9) or rejection marking (10). Contract management device (1) according to one of the Claims 1 until 6 , characterized in that the contract generation module (5) is further configured to transmit a revised contract document (7), in particular a revised contract document (7*) provided with a release mark (9) by an evaluation module (8), to a user (11). Contract management device (1) according to one of the Claims 1 until 7 , characterized in that the contract generation module (5) is further configured to store a revised contract document (7) in the legal database (3) and to interpret this as amended requirements. Method for training a contract management device (1) according to one of the Claims 2 until 8 usable first artificial intelligence (KI1), characterized by the following method steps: - generating the revised contract document (7) by means of the first artificial intelligence (KI1); - processing the revised contract document (7) by means of a second artificial intelligence (KI2), wherein a plurality of training contract documents (12) manually provided with an approval mark (9) or a rejection mark (10) were fed to the second artificial intelligence (KI2) as ground truth in an initial training; - providing the revised contract document (7) with an approval mark (9) or a rejection mark (10) by the second artificial intelligence (KI2); and - feeding the marked revised contract document (7*) to the first artificial intelligence (KI1) as ground truth.