US-20260127535-A1 - SYSTEMS AND METHODS FOR ACCELERATING PROCESS TRANSFORMATION OPPORTUNITY IDENTIFICATION USING GENERATIVE AI

US20260127535A1US 20260127535 A1US20260127535 A1US 20260127535A1US-20260127535-A1

Abstract

Systems and methods for accelerating the identification of process transformation opportunities offer a data-driven, systematic approaches that incorporate generative AI and user input. Organizational processes are modeled to extract customer pain points, create process activity graphs, identify, and compute key performance indicators (KPIs), benchmark against similar organizations, and simulate business value to highlight transformation opportunities. Embodiments leverage existing knowledge base of customer profiles and solutions to enhance the accuracy of process improvement recommendations.

Inventors

Dipanjan Ghosh
Ahmed FARAHAT
Chetan GUPTA

Assignees

HITACHI, LTD.

Dates

Publication Date: 20260507
Application Date: 20241101

Claims (20)

1 . An optimization method for identifying and recommending process transformation opportunities for an organization, the method comprising: providing unstructured data associated with customer information, which is obtained from unstructured data sources comprising at least one of user-entered data, a customer portal, or an external source, as an input to a generative artificial intelligence (AI) to obtain customer attributes; creating a customer profile and storing it, along with embeddings of the customer attributes, in a customer knowledge base (KB), wherein the generative AI uses a large language model to obtain from the unstructured data a customer pain point that is stored as raw data in the customer process KB and as embedded pain point in a customer process vector database (DB); using the embedded pain point and the customer attribute to perform a lookup or a similarity search in the customer process vector DB to identify and obtain a related process activity graph associated with existing customer data and at least one of a related standardized customer pain point or a related process; identifying an activity associated with a node or an edge of the related process activity graph that increases a likelihood of maximizing a value; for at least some of the nodes and/or edges, iteratively performing steps comprising: in response to identifying a key performance indicator (KPI) associated with the activity, computing the KPI by using at least one of the similarity search or a domain expertise and at least one of a formula or KPI information, which is obtained by using a similarity search on the customer KB, to identify a potential improvement in the KPI based on a comparative analysis of the computed KPIs with similar KPIs associated with similar organizations; and selecting and outputting at least one of the customer pain point, process, or a process activity, which based on the computed KPI that has the highest likelihood of maximizing the value.
2 . The method of claim 1 , further comprising, in response to no related standardized customer pain point or related process associated with existing customer data being identified, accepting as related process a user-entered process or domain expert-provided process, and storing the related process as embedded related process in the customer process vector DB.
3 . The method of claim 1 , further comprising, in response to the KPI being identified based on the domain expertise, storing the KPI as new information in the KB.
4 . The method of claim 3 , wherein the similar KPIs are obtained by benchmarking against at least one of organizational data, an industry standard, or publicly available information.
5 . The method of claim 1 , further comprising, in response to no related process activity graph being identified, providing the unstructured data to the generative AI to generate a process activity graph in lieu of the related process activity graph.
6 . The method of claim 1 , wherein the potential improvement is derived from a domain expertise, a business simulation, or automatically by using the formula.
7 . The method of claim 6 , wherein the formula and the KPI information are derived from the domain expertise and are stored in a database.
8 . The method of claim 6 , wherein the formula is argmax i,,k (V) and V corresponds to the value.
9 . The method of claim 1 , wherein the potential improvement serves as an input to a value computation that, for each potential KPI improvement, computes the value.
10 . The method of claim 1 , further comprising, causing the generative AI to use the related process to recommend or generate, based on the customer attributes and the customer pain point, the process activity graph.
11 . The method of claim 1 , further comprising, in response to the value being computed, using the formula to identify at least one of a pain point or a process activity.
12 . The method of claim 1 , further comprising verifying the process activity graph by domain experts to ensure accuracy and saving the verified process activity graph in the process activity vector DB.
13 . The method of claim 1 , further comprising saving the raw data in a process activity graph DB.
14 . The method of claim 1 , wherein the unstructured data sources comprise at least one of a customer discussion, meeting notes, or a recorded meeting.
15 . The method of claim 1 , wherein the customer attributes include a name, an industry, an organization type, and a revenue amount.
16 . The method of claim 1 , wherein the node represents an entity comprising a machine or a material.
17 . The method of claim 1 , further comprising saving a newly identified KPI or process activity in a future use of the customer process vector DB.
18 . The method of claim 1 , further comprising updating the customer KB with pre- and post-transformation KPIs for future comparative analysis.
19 . The method of claim 1 , further comprising encoding and storing, for the related process, a process name in at least one of the customer process vector DB or the customer process KB.
20 . A non-transitory computer-readable medium for storing instructions for executing a process, the instructions comprising: providing unstructured data associated with customer information, which is obtained from unstructured data sources comprising at least one of user-entered data, a customer portal, or an external source, as an input to a generative artificial intelligence (AI) to obtain customer attributes; creating a customer profile and storing it, along with embeddings of the customer attributes, in a customer knowledge base (KB), wherein the generative AI uses a large language model to obtain from the unstructured data a customer pain point that is stored as raw data in the customer process KB and as embedded pain point in a customer process vector database (DB); using the embedded pain point and the customer attribute to perform a lookup or a similarity search in the customer process vector DB to identify and obtain a related process activity graph associated with existing customer data and at least one of a related standardized customer pain point or a related process; identifying an activity associated with a node or an edge of the related process activity graph that increases a likelihood of maximizing a value; for at least some of the nodes and/or edges, iteratively performing steps comprising: in response to identifying a key performance indicator (KPI) associated with the activity, computing the KPI by using at least one of the similarity search or a domain expertise and at least one of a formula or KPI information, which is obtained by using a similarity search on the customer KB, to identify a potential improvement in the KPI based on a comparative analysis of the computed KPIs with similar KPIs associated with similar organizations; and selecting and outputting at least one of the customer pain point, process, or a process activity, which based on the computed KPI that has the highest likelihood of maximizing the value.

Description

BACKGROUND Field The present disclosure is generally directed to information handling systems, and more specifically, to systems and methods for identifying and improving processes in businesses and other entities. Related Art Organizations are composed of various processes, which consist of a network of associated entities and activities. To improve an organization, these processes need to be created or transformed. Traditionally, consultants and domain experts identify opportunities within a process using their experience. However, their perspective can be biased and may overlook other valuable opportunities for transformation. Relying solely on consultants and domain experts alone is non-scalable and introduces the potential of subjective decision-making. Business process mining has been an active area of work and several tools exist. However, these tools seldom consider process information in an unstructured format, as they mostly work on transactional data. To address various technical problems, embodiments herein provide data-driven systems and methods that combine generative artificial intelligence (AI)-based process understanding and human-in-loop approaches to systematically perform some or all of the following steps, which may leverage a knowledge bank of solutions developed based on existing customers: (1) providing systems and methods for modeling organizational process using generative AI and standardized customer profiles comprising extracted attributes; (2) extracting customer pain points; (3) implementing methods for understanding processes and extract processes as graph data structures, where entities and activities are associated with each other; (4) developing methods for process-activity key performance indicator (KPI) identification, computation, and opportunity identification using on-field generated data; and (5) establishing methods for identifying opportunities within the process for transformation by benchmarking against similar organizations and process activities' KPIs, simulating transformation business value, and identifying opportunities worth transforming. SUMMARY In some aspects of the disclosure, an optimization method for identifying and recommending process transformation opportunities for an organization comprises: providing unstructured data associated with customer information, which is obtained from unstructured data sources including at least one of user-entered data, a customer portal, or an external source, as an input to a generative AI to obtain customer attributes; creating a customer profile and storing it, along with embeddings of the customer attributes, in a customer knowledge base (KB), wherein the generative AI uses a large language model (LLM) to obtain from the unstructured data a customer pain point that is stored as raw data in the customer process KB and as embedded pain point in a customer process vector database (DB); using the embedded pain point and the customer attribute to perform a lookup or a similarity search in the customer process vector DB to identify and obtain a related process activity graph associated with existing customer data and at least one of a related standardized customer pain point or a related process; identifying an activity associated with a node or an edge of the related process activity graph that increases a likelihood of maximizing a value; for at least some of the nodes and/or edges, iteratively performing steps including: in response to identifying a KPI associated with the activity, computing the KPI by using at least one of the similarity search or a domain expertise and at least one of a formula or KPI information, which is obtained by using a similarity search on the customer KB, to identify a potential improvement in the KPI based on a comparative analysis of the computed KPIs with similar KPIs associated with similar organizations; and selecting and outputting at least one of the customer pain point, process, or a process activity, which based on the computed KPI that has the highest likelihood of maximizing the value. In some aspects, the optimization method further comprises, in response to no related standardized customer pain point or related process associated with existing customer data being identified, accepting as related process a user-entered process or domain expert-provided process, and storing the related process as embedded related process in the customer process vector DB. In some aspects, the optimization method further comprises, in response to the KPI being identified based on the domain expertise, storing the KPI as new information in the KB. In some aspects, the similar KPIs are obtained by benchmarking against at least one of organizational data, an industry standard, or publicly available information. In some aspects, the optimization method further comprises, in response to no related process activity graph being identified, providing the unstructured data to the generative AI to generate