Automatic product improvement systems and methods via generative artificial intelligence and customer interactions

1 . A system, comprising:

a processor that executes computer-executable instructions stored in a non-transitory computer-readable memory, wherein execution of the computer-executable instructions causes the processor to:

access a deficiency description, wherein the deficiency description is natural language text describing a software deficiency afflicting a software application;

access a software feature repository comprising a plurality of software features each being a respective plug-in, add-on, or functionality that is already coded for, developed for, or part of the software application, wherein each of the plurality of software features is described by a respective textual description;

determine whether available software features in the software feature repository address or solve the software deficiency, wherein such determining comprises at least one of:

comparing first embeddings respectively produced by a generative model for textual descriptions in the software feature repository to an embedding produced by the generative model for the deficiency description; or

identifying textual descriptions in the software feature repository that recite one or more first keywords that are present in the deficiency description;

feed, in response to determining that there is an available software feature in the software feature repository that addresses or solves the software deficiency, to an input layer of the generative model the deficiency description, a textual description corresponding to the available software feature, and a tutorial prompt instructing the generative model to synthesize a text string, a diagram, or a video teaching a user of the software application how to activate or utilize the available software feature, wherein the deficiency description, the textual description, and the tutorial prompt complete a first forward pass through hidden layers of the generative model, and wherein an output layer of the generative model synthesizes as first output the text string, the diagram, or the video based on first hidden activation maps produced by the hidden layers of the generative model;

search, in response to determining that there is no available software feature in the software feature repository that addresses or solves the software deficiency, a technical document repository for one or more technical documents that are relevant to addressing or solving the software deficiency, wherein such searching comprises at least one of:

comparing second embeddings respectively produced by the generative model for technical documents in the technical document repository to the embedding produced by the generative model for the deficiency description; or

identifying technical documents in the technical document repository that recite one or more second keywords that are present in the deficiency description; and

feed to the input layer of the generative model the deficiency description, the one or more technical documents, and a solution generation prompt instructing the generative model to identify, by referring to the one or more technical documents, how to redesign or alter the software application to fix the software deficiency, wherein the deficiency description, the one or more technical documents, and the solution generation prompt complete a second forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as second output a recommended design alteration to the software application based on second hidden activation maps produced by the hidden layers of the generative model.

2 . The system of claim 1 , wherein:

the recommended design alteration comprises synthetic text or synthetic diagrams describing or showing how a code base of the software application should be redesigned to address or solve the software deficiency, and wherein the processor transmits, to a computing device associated with a technician of the software application, an electronic message instructing that the code base be edited in accordance with the recommended design alteration; or

the recommended design alteration comprises one or more synthesized lines of computer code an insertion of which into the code base of the software application would address or solve the software deficiency, and wherein the processor inserts the one or more synthesized lines of computer code into the code base.

3 . The system of claim 1 ;

wherein execution of the computer-executable instructions further causes the processor to determine whether a past electronic client interaction record pertains to the software deficiency;

wherein the searching is in response to both:

determining that there is no available software feature in the software feature repository that addresses or solves the software deficiency; and

determining that the past electronic client interaction record pertains to the software deficiency; and

wherein the past electronic client interaction record comprises at least one of:

a textual conversation between a computing device of a client and a chatbot associated with the software application;

an audio recording captured by a microphone of the computing device and containing verbal questions, complaints, or exclamations spoken by the client during use of the software application;

a video recording captured by a camera of the computing device and showing facial expressions or body language of the client during use of the software application;

a user-interface tracking log captured by the computing device and showing where the client has clicked, moved, or scrolled within a user-interface of the software application; or

a textual document typed on the computing device and pertaining to the software application.

4 . The system of claim 3 , wherein the determining whether the past electronic client interaction record pertains to the software deficiency comprises:

feeding to the input layer of the generative model the deficiency description, the past electronic client interaction record, and a common deficiency prompt instructing the generative model to determine whether or not the past electronic client interaction indicates or suggests the same deficiency as the deficiency description, wherein the deficiency description, the past electronic client interaction record, and the common deficiency prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the common deficiency prompt based on third hidden activation maps produced by the hidden layers of the generative model.

5 . The system of claim 1 :

wherein the comparing the first embeddings or the identifying the textual descriptions reveals a potentially-relevant available software feature in the software feature repository that might potentially address or solve the software deficiency; and

wherein execution of the computer-executable instructions further causes the processor to:

feed to the input layer of the generative model the deficiency description, another textual description of the potentially-relevant software feature, and a helpfulness prompt instructing the generative model to determine whether or not the potentially-relevant software feature would help to address or solve the software deficiency, wherein the deficiency description, the another textual description, and the helpfulness prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the helpfulness prompt based on third hidden activation maps produced by the hidden layers of the generative model.

6 . The system of claim 1 , wherein execution of the computer-executable instructions further causes the processor to:

access feedback indicating a client's audible reaction or body language reaction to the recommended design alteration; and

feed to the input layer of the generative model the recommended design alteration, the feedback, and an acceptability prompt instructing the generative model to determine whether or not the feedback indicates or suggests that the recommended design alteration is acceptable to the client, wherein the recommended design alteration, the feedback, and the acceptability prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the acceptability prompt based on third hidden activation maps produced by the hidden layers of the generative model.

7 . The system of claim 1 , wherein the solution generation prompt further instructs the generative model to:

identify, by referring to the one or more technical documents, a personnel role at a manufacturer of the software application to whose attention the software deficiency should be brought; and

tailor a text-image-code proportion of the recommended design alteration to the personnel role.

8 . The system of claim 1 , wherein the recommended design alteration comprises reference indicators that point to which of the one or more technical documents that the generative model relied upon to synthesize the recommended design alteration.

9 . The system of claim 1 , wherein the one or more technical documents that are relevant to addressing or solving the software deficiency comprise:

chapters or sections of an operating manual of the software application;

service log entries recorded by the software application;

standard operating procedures promulgated by a manufacturer of the software application; or

sections of a code base of the software application.

10 . A computer-implemented method, comprising:

accessing, by a device operatively coupled to a processor, a deficiency description, wherein the deficiency description is natural language text describing a software deficiency afflicting a software application;

accessing, by the device, a software feature repository comprising a plurality of software features each being a respective plug-in, add-on, or functionality that is already coded for, developed for, or part of the software application, wherein each of the plurality of software features is described by a respective textual description;

determining, by the device, whether available software features in the software feature repository address or solve the software deficiency, wherein such determining comprises at least one of:

comparing first embeddings respectively produced by a generative model for textual descriptions in the software feature repository to an embedding produced by the generative model for the deficiency description; or

identifying textual descriptions in the software feature repository that recite one or more first keywords that are present in the deficiency description;

feeding, by the device and in response to determining that there is an available software feature in the software feature repository that addresses or solves the software deficiency, to an input layer of the generative model the deficiency description, a textual description corresponding to the available software feature, and a tutorial prompt instructing the generative model to synthesize a text string, a diagram, or a video teaching a user of the software application how to activate or utilize the available software feature, wherein the deficiency description, the textual description, and the tutorial prompt complete a first forward pass through hidden layers of the generative model, and wherein an output layer of the generative model synthesizes as first output the text string, the diagram, or the video based on first hidden activation maps produced by the hidden layers of the generative model

searching, by the device and in response to determining that there is no available software feature in the software feature repository that addresses or solves the software deficiency, a technical document repository for one or more technical documents that are relevant to addressing or solving the software deficiency, wherein such searching comprises at least one of:

comparing second embeddings respectively produced by the generative model for technical documents in the technical document repository to the embedding produced by the generative model for the deficiency description; or

identifying technical documents in the technical document repository that recite one or more second keywords that are present in the deficiency description; and

feeding, by the device, to the input layer of the generative model the deficiency description, the one or more technical documents, and a solution generation prompt instructing the generative model to identify, by referring to the one or more technical documents, how to redesign or alter the software application to fix the software deficiency, wherein the deficiency description, the one or more technical documents, and the solution generation prompt complete a second forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as first output a recommended design alteration to the software application based on second hidden activation maps produced by the hidden layers of the generative model.

11 . The computer-implemented method of claim 10 , wherein:

the recommended design alteration comprises synthetic text or synthetic diagrams describing or showing how a code base of the software application should be redesigned to address or solve the software deficiency, and further comprising transmitting, by the device and to a computing device associated with a technician of the software application, an electronic message instructing that the code base be edited in accordance with the recommended design alteration; or

the recommended design alteration comprises one or more synthesized lines of computer code an insertion of which into the code base of the software application would address or solve the software deficiency, and further comprising inserting, by the device, the one or more synthesized lines of computer code into the code base.

12 . The computer-implemented method of claim 10 :

wherein the computer-implemented method further comprises determining, by the device, whether a past electronic client interaction record pertains to the software deficiency;

wherein the searching is in response to both:

determining that there is no available software feature in the software feature repository that addresses or solves the software deficiency; and

determining that the past electronic client interaction record pertains to the software deficiency; and

wherein the past electronic client interaction record comprises at least one of:

a textual conversation between a computing device of a client and a chatbot associated with the software application;

an audio recording captured by a microphone of the computing device and containing verbal questions, complaints, or exclamations spoken by the client during use of the software application;

a video recording captured by a camera of the computing device and showing facial expressions or body language of the client during use of the software application;

a user-interface tracking log captured by the computing device and showing where the client has clicked, moved, or scrolled within a user-interface of the software application; or

a textual document typed on the computing device and pertaining to the software application.

13 . The computer-implemented method of claim 12 , wherein the determining whether the past electronic client interaction record pertains to the software deficiency comprises:

feeding to the input layer of the generative model the deficiency description, the past electronic client interaction record, and a common deficiency prompt instructing the generative model to determine whether or not the past electronic client interaction indicates or suggests the same deficiency as the deficiency description, wherein the deficiency description, the past electronic client interaction record, and the common deficiency prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the common deficiency prompt based on third hidden activation maps produced by the hidden layers of the generative model.

14 . The computer-implemented method of claim 10 :

wherein the comparing the first embeddings or the identifying the textual descriptions reveals a potentially-relevant available software feature in the software feature repository that might potentially address or solve the software deficiency; and

wherein the computer-implemented method further comprises feeding, by the device, to the input layer of the generative model the deficiency description, another textual description of the potentially-relevant software feature, and a helpfulness prompt instructing the generative model to determine whether or not the potentially-relevant software feature would help to address or solve the software deficiency, wherein the deficiency description, the another textual description, and the helpfulness prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the helpfulness prompt based on third hidden activation maps produced by the hidden layers of the generative model.

15 . The computer-implemented method of claim 10 , further comprising:

accessing, by the device, feedback indicating a client's audible reaction or body language reaction to the recommended design alteration; and

feeding, by the device, to the input layer of the generative model the recommended design alteration, the feedback, and an acceptability prompt instructing the generative model to determine whether or not the feedback indicates or suggests that the recommended design alteration is acceptable to the client, wherein the recommended design alteration, the feedback and the acceptability prompt complete a third forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as third output text answering the acceptability prompt based on third hidden activation maps produced by the hidden layers of the generative model.

16 . The computer-implemented method of claim 10 , wherein the solution generation prompt further instructs the generative model to:

identify, by referring to the one or more technical documents, a personnel role at a manufacturer of the software application to whose attention the software deficiency should be brought; and

tailor a text-image-code proportion of the recommended design alteration to the personnel role.

17 . The computer-implemented method of claim 10 , wherein the recommended design alteration comprises reference indicators that point to which of the one or more technical documents that the generative model relied upon to synthesize the recommended design alteration.

18 . The computer-implemented method of claim 10 , wherein the one or more technical documents that are relevant to addressing or solving the software deficiency comprise:

chapters or sections of an operating manual of the software application;

service log entries recorded by the software application;

standard operating procedures promulgated by a manufacturer of the software application; or

sections of a code base of the software application.

19 . A computer program product for facilitating client-tailored software application alteration via generative models, the computer program product comprising a non-transitory computer-readable memory having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

access a deficiency description, wherein the deficiency description is natural language text describing a software deficiency afflicting a software application;

access a software feature repository comprising a plurality of software features each being a respective plug-in, add-on, or functionality that is already coded for, developed for, or part of the software application, wherein each of the plurality of software features is described by a respective textual description;

determine whether available software features in the software feature repository address or solve the software deficiency, wherein such determining comprises at least one of:

comparing first embeddings respectively produced by a generative model for textual descriptions in the software feature repository to an embedding produced by the generative model for the deficiency description; or

identifying textual descriptions in the software feature repository that recite one or more first keywords that are present in the deficiency description;

feed, in response to determining that there is an available software feature in the software feature repository that addresses or solves the software deficiency, to an input layer of the generative model the deficiency description, a textual description corresponding to the available software feature, and a tutorial prompt instructing the generative model to synthesize a text string, a diagram, or a video teaching a user of the software application how to activate or utilize the available software feature, wherein the deficiency description, the textual description, and the tutorial prompt complete a first forward pass through hidden layers of the generative model, and wherein an output layer of the generative model synthesizes as first output the text string, the diagram, or the video based on first hidden activation maps produced by the hidden layers of the generative model

search, in response to determining that there is no available software feature in the software feature repository that addresses or solves the software deficiency, a technical document repository for one or more technical documents that are relevant to addressing or solving the software deficiency, wherein such searching comprises at least one of:

comparing second embeddings respectively produced by the generative model for technical documents in the technical document repository to the embedding produced by the generative model for the deficiency description; or

identifying technical documents in the technical document repository that recite one or more second keywords that are present in the deficiency description;

feed to the input layer of the generative model the deficiency description, the one or more technical documents, and a solution generation prompt instructing the generative model to identify, by referring to the one or more technical documents, how to redesign or alter the software application to fix the software deficiency, wherein the deficiency description, the one or more technical documents, and the solution generation prompt complete a second forward pass through the hidden layers of the generative model, and wherein the output layer of the generative model synthesizes as second output a recommended design alteration to the software application based on second hidden activation maps produced by the hidden layers of the generative model, wherein the recommended design alteration comprises one or more synthesized lines of computer code an insertion of which into a code base of the software application would address or solve the software deficiency; and

insert the one or more synthesized lines of computer code into the code base of the software application.

20 . The computer program product of claim 19 , wherein the generative model creates the deficiency description based on being executed on an electronic client interaction record that comprises at least one of:

a textual conversation between a computing device of a client and a chatbot associated with the software application;

an audio recording captured by a microphone of the computing device and containing verbal questions, complaints, or exclamations spoken by the client during use of the software application;

a video recording captured by a camera of the computing device and showing facial expressions or body language of the client during use of the software application;

a user-interface tracking log captured by the computing device and showing where the client has clicked, moved, or scrolled within a user-interface of the software application; or

a textual document typed on the computing device and pertaining to the software application.