Please consider supporting us by disabling your content blocker.
loader

Twenty-first century manufacturers post-COVID-19 have been facing significant challenges across their functions in supply chain, risk, operations, and customer experience. Threats by new (often more technologically capable) entrants, competition consolidation, resource sustainability and changing regulations add to the mix. At the same time, one cannot avoid conversations discussing the rapid advancements in technology across sectors that are often touted as the saviors of manufacturing. If you find yourself in such conversations, it’s good to be aware of what lies on the horizon, what is truly applicable and what is mostly hype. Tackling the problems of today with the technology of tomorrow will require reimagining people, processes, and systems.

Today’s environment must shift from People executing Processes, presented with Data, powered by Technology

to …

Technology powered by Data executing Processes managed by People

What is Decision Intelligence?

Much has already been said about artificial intelligence, but what about Decision Intelligence? For the purposes of this discussion, the authors offer the following definition:

Decision Intelligence is the efficient enablement of insights driven to the point of action by humans or machines.

“Efficient enablement” implies there is an accelerated pace of proceeding from insights to action with as little waste of time, energy, or resources as possible. The “point of action” is where value is created or destroyed, and this is largely based on either the quality of the insights driving the action, or the timeliness of the action being taken. The systems and devices for capturing data, the models for analysis, the workflow for delivering information, and the process for governing and automating actions all fall under the umbrella of Decision Intelligence.

To help illustrate this point, let us consider the myriad use cases for Decision Intelligence in guiding manufacturers toward achieving an autonomous supply chain by simultaneously enhancing the levels of automation, collaboration, and intelligence within the operations. There are at least four tiers to this stepwise progression in capabilities, as shown in Figure 1.

Figure 1: The journey to the autonomous supply chain

The first level is that of functional visibility, which involves the capability to review and react to dashboards and reports. This tends to be single-functional to cross-functional in nature.

Cross-functional and proactive exception management is enabled when these dashboards/reports proactively highlight potential issues as triggers for further escalation.

The third level is end-to-end orchestration, where changes in the supply chain take place and alternative scenarios are developed to determine the optimal solution to issues in near-real time.

The final echelon is the autonomous supply chain, wherein the system identifies and executes allowable decisions autonomously while also escalating complex issues to the attention of users with supporting information and recommendations for decision-making.

What is Causal AI?

Decision Intelligence is not possible without the underlying decision models. Until recently, these models mostly consisted of data models (powered by statistical, AI/ML algorithms) designed to seek the most optimal solutions linking contributing factors (features) to desired measurable outcomes (targets). Depending on the technique, the connection between cause and effect is not always clear. In fact, it is often difficult to distinguish between correlation and causation. This opacity is often unacceptable in manufacturing decision-making and has been one of the main hindrances to at-scale adoption by business users.

Another technological development, known as Causal AI, counters the effects of the black box of typical data models by pairing such models with knowledge models based on domain expertise. With the use of a causal model as its basis, the Causal AI technique makes inferences using causality rather than just correlation. Organizations are therefore able to use this methodology to explain decision-making through the lens of the causes leading to specific outcomes.

One of the approaches used in Causal AI, based on long-understood principles, is known as causal graph modeling. A Causal graph is a pictorial graph made up of collections of variables (nodes), linked by arrows (edges) to show the cause-and-effect relationships between them. There are two ways to determine a causal graph: 1) expert domain knowledge and 2) causal discovery algorithms. We will focus on the former for this manufacturing application.

For Causal AI to work in Decision Intelligence, knowledge models are created to help guide the data model by applying constraints and queries. The knowledge model can be built from any number of applicable sources of domain expertise, including human experts, theories of failure and operation, equipment, and process manuals. Ingestion is very efficiently enabled by Small Language Models (SLMs), a sub-technique of Generative AI. Another advancement in Generative AI is Large Language Models (LLMs). LLMs generate, extract, match, cluster, classify, summarize, and rewrite text as if it has been written by a human. This can be used to generate the explanation of the output from the combined decision model in understandable human and business terms.

There are at least three reasons why now is the time for Decision Intelligence powered by Causal AI in manufacturing:

  1. The advances in the field of AI to ingest domain expertise at scale into knowledge models.
  2. The increasing pervasiveness of large-scale manufacturing data at high-speed becoming readily available.
  3. The burning platform for manufacturers to differentiate themselves based on the speed of decision-making in a highly competitive environment.

Four value drivers for scaling a Causal AI Decision Intelligence platform

Organizations across industries are navigating complex supply chain challenges to drive the next wave of business growth. In response to this challenge, the EY team developed an AI-enabled digital platform for driving enterprise supply chain transformations that is already live at more than 30 global clients. The core value for these clients comes from maintaining one platform which incorporates cutting edge AI/ML, Gen AI and optimization engines for a portfolio of solutions seamlessly integrated with and not replacing their legacy source systems.

As a company gets started on the Decision Intelligence path, one should consider implementing a similar unified platform that will enable analytics-led decision-making through visibility, predictivity, accountability, and governance. These solutions will center around four key value drivers for a leap in performance for the supply chain and operations: 1) Supply, 2) Demand, 3) Sales, 4) Costs.

Figure 2: Four value drivers for the Decision Intelligence Platform

The views expressed are those of the authors and do not necessarily represent the views of Ernst & Young LLP or any other member firm of the global EY organization.

About the authors:

Travis Wolf is a partner/principal in EY’s Supply Chain practice, focused on helping our clients achieve a resilient and flexible supply chain through the power of process and digital transformation.

Carl B. March is a Manufacturing Transformation Leader at EY. He works with manufacturers to architect, accelerate, execute and scale their most ambitious manufacturing transformation visions.

SCMR

Source: link