Using generative AI to make a natural language interface
An operations computing system receives, from a user computing device, initial natural language text input associated with an incident and generates, based on the initial natural language text input, a set of prompts. The operations computing system provides the set of prompts as input to a machine learning model and receives, from the model, text output for each prompt, in which the text output includes a clarifying question or a clarifying instruction. The operations computing system sends, to the user computing device, the text output, and receives, from the user computing device, additional natural language text input. The operations computing system applies the model to the natural language text input to generate respective initial structured text data for each prompt. The operations computing system applies the model to the respective initial structured text data for each prompt to generate updated structured text data including instructions for creating an incident workflow.
1 . A method comprising:
receiving, by a computing system, and from a user computing device, an initial natural language text input associated with an incident;
identifying, by the computing system, based on the initial natural language text input, one or more tasks;
generating, by the computing system and based on the initial natural language text input, a set of prompts including one or more prompts, wherein each respective prompt from the one or more prompts is associated with a respective task of the one or more tasks, wherein the respective prompt includes a description of the respective task, and wherein the respective prompt includes one or more additional instructions instructing a machine learning model to output a clarifying question or a clarifying instruction if the machine learning model does not have information associated with the respective task;
for each respective prompt from the one or more prompts:
providing, by the computing system, the respective prompt as input to the machine learning model;
receiving, by the computing system and from the machine learning model, a respective text output for the respective prompt, wherein the respective text output includes one or more of the clarifying question or the clarifying instruction;
sending, by the computing system and to the user computing device, the respective text output;
receiving, by the computing system, and from the user computing device, a respective additional natural language text input for the respective prompt;
providing, by the computing system, the initial natural language text input and the respective additional natural language text input for the respective prompt as input to the machine learning model; and
obtaining, by the computing system, respective initial structured text data for the respective prompt from the machine learning model;
providing, by the computing system, all respective initial structured text data for the one or more prompts as input to the machine learning model; and
obtaining, by the computing system, updated structured text data for the one or more prompts from the machine learning model, wherein the updated structured text data for the one or more prompts includes instructions for creating an incident workflow for the incident.
2 . The method of claim 1 , further comprising:
receiving, by the computing system, the updated structured text data for the one or more prompts; and
creating, by the computing system, and based on the updated structured text data for the one or more prompts, the incident workflow.
3 . The method of claim 1 , further comprising,
generating, by the computing system, data for a user interface including the initial natural language text input, the respective additional natural language text input, and the respective text output.
4 . The method of claim 1 , wherein the initial natural language text input and the respective additional natural language text input are indicative of data for the incident workflow.
5 . The method of claim 1 , further comprising:
determining, by the computing system and based on stored data, whether the initial natural language text input and the respective additional natural language text input are valid, wherein the stored data includes one or more of valid user data, valid computing system data, and valid workflow parameters data.
6 . The method of claim 5 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include one or more of a trigger, an action, and a field for the incident workflow.
7 . The method of claim 5 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include a trigger, and wherein the trigger includes one or more of a manual trigger, an automatic trigger, a scheduled trigger, and an event-based trigger.
8 . The method of claim 5 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include an action, and wherein the action includes one or more of an addition of stakeholders, a sending of a status update, a creation of a message thread, a sending of a message thread link, an addition of responders, and a starting of a virtual meeting.
9 . The method of claim 5 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include a field for the incident workflow, and wherein the field for the incident workflow includes one or more of an identifier, a title, a description, a timestamp, an incident type, an incident source, a severity level, one or more assigned users, an urgency level, a priority level, a current incident status, incident resolution data, one or more associated support tickets, and an action log.
10 . The method of claim 1 , further comprising:
receiving, by the computing system, user input as feedback data for the incident workflow.
11 . A system comprising:
a memory; and
one or more processors having access to the memory, wherein the one or more processors are configured to:
receive, from a user computing device, an initial natural language text input associated with an incident;
identify, based on the initial natural language text input, one or more tasks;
generate, based on the initial natural language text input, a set of prompts including one or more prompts, wherein each respective prompt from the one or more prompts is associated with a respective task of the one or more tasks, wherein the respective prompt includes a description of the respective task, and wherein the respective prompt includes one or more additional instructions instructing a machine learning model to output a clarifying question or a clarifying instruction if the machine learning model does not have information associated with the respective task;
for each respective prompt from the one or more prompts:
provide the respective prompt as input to the machine learning model;
receive, from the machine learning model, a respective text output for the respective prompt, wherein the respective text output includes one or more of the clarifying question or the clarifying instruction;
send, to the user computing device, the respective text output;
receive, from the user computing device, a respective additional natural language text input for the respective prompt;
provide the initial natural language text input and the respective additional natural language text input for the respective prompt as input to the machine learning model; and
obtain respective initial structured text data for the respective prompt from the machine learning model;
provide all respective initial structured text data for the one or more prompts as input to the machine learning model; and
obtain updated structured text data for the one or more prompts from the machine learning model, wherein the updated structured text data for the one or more prompts includes instructions for creating an incident workflow for the incident.
12 . The system of claim 11 , wherein the one or more processors are further configured to:
receive the updated structured text data for the one or more prompts; and
create, based on the updated structured text data for the one or more prompts, the incident workflow.
13 . The system of claim 11 , wherein the one or more processors are further configured to generate data for a user interface including the initial natural language text input, the respective additional natural language text input, and the respective text output.
14 . The system of claim 11 , wherein the initial natural language text input and the respective additional natural language text input are indicative of data for the incident workflow.
15 . The system of claim 11 , wherein the one or more processors are further configured to:
determine, based on stored data, whether the initial natural language text input and the respective additional natural language text input are valid, wherein the stored data includes one or more of valid user data, valid computing system data, and valid workflow parameters data.
16 . The system of claim 15 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include one or more of a trigger, an action, and a field for the incident workflow, wherein the trigger includes one or more of a manual trigger, an automatic trigger, a scheduled trigger, and an event-based trigger, wherein the action includes one or more of an addition of stakeholders, a sending of a status update, a creation of a message thread, a sending of a message thread link, an addition of responders, and a starting of a virtual meeting, and wherein the field for the incident workflow includes one or more of an identifier, a title, a description, a timestamp, an incident type, an incident source, a severity level, one or more assigned users, an urgency level, a priority level, a current incident status, incident resolution data, one or more associated support tickets, and an action log.
17 . A non-transitory computer-readable storage medium encoded with instructions that, when executed, cause at least one processor of a computing system to:
receive, from a user computing device, an initial natural language text input associated with an incident;
identify, based on the initial natural language text input, one or more tasks;
generate, based on the initial natural language text input, a set of prompts including one or more prompts, wherein each respective prompt from the one or more prompts is associated with a respective task of the one or more tasks, wherein the respective prompt includes a description of the respective task, and wherein the respective prompt includes one or more additional instructions instructing a machine learning model to output a clarifying question or a clarifying instruction if the machine learning model does not have information associated with the respective task;
for each respective prompt from the one or more prompts:
provide the respective prompt as input to the machine learning model;
receive, from the machine learning model, a respective text output for the respective prompt, wherein the respective text output includes one or more of the clarifying question or the clarifying instruction;
send, to the user computing device, the respective text output;
receive, from the user computing device, a respective additional natural language text input for the respective prompt;
provide the initial natural language text input and the respective additional natural language text input for the respective prompt as input to the machine learning model; and
obtain respective initial structured text data for the respective prompt from the machine learning model;
provide all respective initial structured text data for the one or more prompts as input to the machine learning model;
obtain updated structured text data for the one or more prompts from the machine learning model, wherein the updated structured text data for the one or more prompts includes instructions for creating an incident workflow for the incident;
receive the updated structured text data for the one or more prompts; and
create, based on the updated structured text data for the one or more prompts, the incident workflow.
18 . The non-transitory computer-readable storage medium of claim 17 , wherein the instructions further cause the at least one processor to generate data for a user interface including the initial natural language text input, the respective additional natural language text input, and the respective text output.
19 . The non-transitory computer-readable storage medium of claim 17 , wherein the initial natural language text input and the respective additional natural language text input are indicative of data for the incident workflow, and wherein the instructions further cause the at least one processor:
determine, based on stored data, whether the initial natural language text input and the respective additional natural language text input are valid, wherein the stored data includes one or more of valid user data, valid computing system data, and valid workflow parameters data.
20 . The non-transitory computer-readable storage medium of claim 19 , wherein the initial natural language text input and the respective additional natural language text input determined to be valid include one or more of a trigger, an action, and a field for the incident workflow, wherein the trigger includes one or more of a manual trigger, an automatic trigger, a scheduled trigger, and an event-based trigger, wherein the action includes one or more of an addition of stakeholders, a sending of a status update, a creation of a message thread, a sending of a message thread link, an addition of responders, and a starting of a virtual meeting, and wherein the field for the incident workflow includes one or more of an identifier, a title, a description, a timestamp, an incident type, an incident source, a severity level, one or more assigned users, an urgency level, a priority level, a current incident status, incident resolution data, one or more associated support tickets, and an action log.