In collaboration with

Boston Consulting Group

Harnessing the
AI Revolution in
Industrial Operations:
A Guidebook
WHITE PAPER
OCTOBER 2023
Images: Getty Images

Contents
Foreword 3

Executive summary 4

1. AI’s potential in industrial operations 5

2. Harnessing the AI revolution in industrial operations: a guidebook 9

2.1 Paving the way for success from AI in industrial operations 11

2.2 Mastering the AI journey across production networks 12

2.3 Staying at the forefront of AI innovations 13

2.4 Foundations for AI implementation in industrial 13

operations at scale

Conclusion 17

Contributors 18

Endnotes 20

Disclaimer
This document is published by the
World Economic Forum as a contribution
to a project, insight area or interaction.
The findings, interpretations and
conclusions expressed herein are a result
of a collaborative process facilitated and
endorsed by the World Economic Forum
but whose results do not necessarily
represent the views of the World Economic
Forum, nor the entirety of its Members,
Partners or other stakeholders.

© 2023 World Economic Forum. All rights

reserved. No part of this publication may
be reproduced or transmitted in any form
or by any means, including photocopying
and recording, or by any information
storage and retrieval system.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 2

OCTOBER 2023 Harnessing the AI Revolution in Industrial Operations:
A Guidebook

Foreword
Daniel Küpper Kyriakos Triantafyllidis
Managing Director and Head of Growth and
Senior Partner, Global Strategy, Centre for
(Co-)Lead BCG Manufacturing Advanced Manufacturing
and Supply Chain, Boston and Supply Chains,
Consulting Group World Economic Forum

Although continuous improvements in production support manufacturers on their AI journey, ensuring

efficiency have historically been a cornerstone they harness AI’s full potential in manufacturing and
for maintaining the competitiveness of global supply chains.
enterprises, conventional means of enhancing
productivity in manufacturing are becoming This white paper delves into the initiative’s findings
increasingly exhausted. On the other hand, over the past year, derived from consultations
the growing significance of data and advanced with a community of leading manufacturers,
manufacturing technologies, such as artificial technology innovators and academic experts. It
intelligence (AI), offers fresh avenues to improve not offers insights into the current state of AI adoption
only production efficiency and flexibility, but also to in industrial operations and sheds light on recent
drive sustainability and empower the workforce. advancements in the field. The centrepiece
of the paper is a guidebook intended to help
Many manufacturers have already integrated AI into manufacturing companies continuously capture
their operations. However, the majority struggle the full value of AI as the technology evolves. The
to achieve their AI-related targets. This shortfall guidebook highlights the crucial considerations
predominantly arises from inadequate organizational and steps of the AI journey – from defining the
and technological foundations, which are essential overarching objectives of AI in operations to
for scaling AI solutions throughout production identifying, building and scaling the relevant
networks. At the same time, AI’s evolution applications and required foundations to ensure
continues unabated, with innovations such as they stay at the forefront of AI innovations.
generative AI emerging on a regular basis.
We trust that this white paper and guidebook
In 2022, the World Economic Forum – in significantly contribute to efforts to deploy AI in
collaboration with Boston Consulting Group (BCG), industrial operations by helping manufacturers
the Centre for the Fourth Industrial Revolution in navigate their AI transformation to reach the next
Türkiye and TÜV SÜD – launched a global initiative frontiers of productivity, agility, sustainability and
on AI-Powered Industrial Operations. The goal is to workforce engagement.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 3

Executive summary
This report presents a guidebook
for harnessing the AI revolution in
industrial operations.

Manufacturing companies operate in an manufacturing company’s AI journey and the latter

increasingly complex environment marked by two describing the building blocks needed for a
heightened economic pressures, the sustainability successful AI implementation and scaling:
imperative, the need for resilience and escalating
capability challenges coupled with a talent – Paving the way for success from AI in
shortage.1 Advanced manufacturing technologies, industrial operations: Highlighting the
including AI, have a pivotal role in enabling importance of articulating an organization’s
companies to navigate these challenges and tap long-term AI objectives and transformation
into new value streams. According to a global principles as a starting point on the AI journey.
BCG study on AI in industrial operations from
2023,2 these technologies enable productivity – Mastering the AI journey across production
enhancements of more than 20%. Recognizing networks: Describing a step-by-step approach
the opportunity, about 90% of companies across from the status quo assessment to the
various industries are looking to incorporate AI design, engineering, implementation and
into their operations. However, this contrasts scaling of value-adding AI applications and
with the fact that only one in six of these required foundations.
companies have met their AI-related objectives
to date. This shortfall is primarily attributed to a – Staying at the forefront of AI innovations:
lack of foundational prerequisites both from an Explaining the importance of conducting
organizational and a technological perspective. periodic AI reviews to continuously identify
and integrate AI innovations and related
Although most industrial companies struggle opportunities as they emerge.
to realize the full potential of AI, the technology
continues to evolve.3 Innovations such as generative – Value-adding AI applications for industrial
AI present additional opportunities to reinvent operations: Pinpointing AI-based applications
certain operational processes and to transform how that address inefficiencies and operational
employees work in plants. For example, through opportunities to achieve intended improvements
generative AI it is possible to provide employees and outcomes.
with question-and-answer platforms or give
detailed work instructions for specific maintenance – Foundations for AI implementation in
incidents, including visualizations, the required industrial operations at scale: Defining the
spare parts and other essential information. AI-related organizational and technological
These capabilities remain largely untapped foundations to enable implementation at scale
but can be successfully adopted with the right and long-lasting success.
implementation approach.
Recognizing that an AI journey is not a one-
To support manufacturing companies on their time effort, the guidebook helps manufacturers
AI journey, the World Economic Forum’s Centre to continually adapt to the rapid advancements
for Advanced Manufacturing and Supply Chains and innovations of AI applications in industrial
– in collaboration with BCG and a community operations.
of operations and technology executives as
well as academic experts – has co-developed Moving forward, the World Economic Forum’s
a guidebook for harnessing the AI revolution in Centre for Advanced Manufacturing and Supply
industrial operations. This guidebook draws upon Chains will continue to work closely with
insights gained by exploring the untapped potential manufacturing stakeholders across industries to
of AI in industrial operations and the variety of AI support the long-term journey towards AI-powered
applications that manufacturers currently deploy. industrial operations by incubating innovative AI
pilots and shedding light on the most common
The guidebook consists of five sections, with the quality gaps observed when implementing AI
first three representing the different stages of a systems in an industrial context.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 4

1 AI’s potential in
industrial operations
Most manufacturers are still unable to
successfully implement and scale AI, yet the
evolution of AI regularly reveals new innovations
and opportunities.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 5

 The operating environment for manufacturing automation. Applications, such as demand
companies has grown increasingly complex, forecasting or computer vision-based quality
characterized by economic pressure, the control, are becoming increasingly prevalent
sustainability imperative, the necessity for agile in manufacturing operations.5 According to a
operations and capability challenges exacerbated recent BCG study involving more than 1,800
by talent shortages.4 Advanced manufacturing manufacturing executives worldwide, 89% of
technologies such as AI can help manufacturers respondents across industries regard AI as
address this complexity and unlock the next wave essential. A notable 68% of manufacturers
of value through improvements in productivity, have already started their AI journey, fully
enhanced production flexibility and workforce implementing at least one AI use case in
support, among other benefits. Seeing the production. However, a mere one-sixth (16%)
opportunities, leading manufacturers are exploring of these companies have already achieved
how to implement these technologies. their AI-related targets (Figure 1). Nearly all
executives (98%) cited the primary reasons for
The adoption of AI technologies such as machine this shortfall as inadequate organizational
learning (ML) and deep learning (DL) has facilitated foundations – such as a lack of digital skills and
a shift from traditional deterministic and rule- capabilities – and a lagging technology base in
based systems to probabilistic and goal-oriented areas such as data processing.6

FIGURE 1: Current state of AI adoption in industrial operations

89%
of executives aim to implement AI in their
production

68%
of manufacturers have already started their AI
journey and have fully implemented at least one
AI use case

16%
of AI adopters have successfully achieved their
AI-related targets

Source: BCG, Global Survey on AI in Industrial Operations 20237

These findings confirm that activities and efforts times greater performance than the previous chip
are underway within industrial operations to generation.8 In this context, recent advancements
adopt AI. Yet many companies have become in generative AI present exciting opportunities,
stuck in the piloting phase and are unable to promising to unlock potentials that extend well
scale their solutions and capture the full value of beyond the use of ChatGPT in the manufacturing
AI across production networks and value chains. environment. Based on consultations with
At the same time, the evolution of AI and related AI experts, the enabled applications can be
technologies continues – an example is the categorized into three main types: assistance
introduction of 4nm chips featuring an astounding systems, recommendation systems and
80 trillion transistors, which can deliver up to 6.7 autonomous systems (Figure 2).

Harnessing the AI Revolution in Industrial Operations: A Guidebook 6

 FIGURE 2: Three main types of generative AI applications in manufacturing

– Workforce questions and answers: Allowing individuals to pose questions – e.g. on

a bill of materials (BOM) – and receive instantaneous answers tailored to the enquirer’s
background

– Machine and programmable logic controller (PLC) programming co-pilot:

Facilitating the programming of machines by drafting code or code blocks or reviewing
Assistance systems existing code – e.g. to automate a manual process
Generative AI applications that
are similar to a ChatGPT tool in – Process report and documentation tool: Enabling the automated generation of
the manufacturing environment, documents such as quality reports after a major product defect, or maintenance
enhancing the efficiency of manual documentation following an equipment breakdown and repair
activities such as programming or
machine maintenance

– Maintenance co-pilot: Extending the ML- or DL-based predictive maintenance

approach with proposed responses. The co-pilot recognizes incidents or alerts from the
predictive systems – e.g. the need to swap out a particular machine component – and
augments them with step-by-step maintenance instructions, accompanied by images
and visualizations, or even guidance on the required spare parts and tools

Recommendation systems – Root cause suggestion tool: Advising how to sustainably solve root causes identified
Generative AI applications making by failure analyses, such as by eliminating recurring issues in an assembly process or
recommendations, thereby helping machine incidents
operators pinpoint the most optimal
solutions for specific tasks – Quality and yield improvement tool: Suggesting, for example, the ideal components
for setpoint optimization via ML in order to increase yield

– Synthetic training data generation: Enabling the synthetic creation of training data
with unprecedented speed and accuracy – e.g. a computer vision-based quality control
process must be trained on a significant number of samples of good and defective
parts to ensure high reliability. These samples are typically collected during the ramp-up
phase of a new product, which initially results in diminished system performance and
necessitates manual intervention. In contrast, using synthetically generated training data
Autonomous systems can markedly accelerate this process
Generative AI applications with
capabilities for self-control and – Robotics transformers: Allowing for the automation of less-repetitive material handling
adaptability to new environments procedures and swift adoption of robotics to unfamiliar environments without the need
for individual retraining and data labelling

Source: World Economic Forum and BCG analysis

The above examples provide an initial glimpse from increased labour efficiency and heightened
into the transformative applications and impact of productivity to cost reductions, as well as
generative AI within the manufacturing environment – augmented production flexibility.

Across multiple industries, companies need to process thousands of shipments daily. SAP has used
generative AI to process goods receipts for these shipments. Delivery details from unstructured paper
documents are automatically extracted and populated into the transportation management system,
which reduces the effort of employees, who otherwise have to read and enter the delivery note data
manually. Implementing this solution at an automotive company made it possible to reduce the receipt
processing time by 55–70%.9

Harnessing the AI Revolution in Industrial Operations: A Guidebook 7

 As of today, many of the generative AI applications Such developments create a pressing need for
described are still in the research stage or have approaches that enable more rapid implementation
been implemented only in pilot projects. Yet, and scaling of innovations. To embrace these AI
considering the rapid technological advancements developments in a target-oriented way, a holistic
and the growing interest generative AI is triggering strategy for AI adoption is needed, encompassing a
within the manufacturing community, it is expected clear vision of AI tailored to the individual industrial
that these solutions will soon be present in industrial context, the suitable value-adding applications and
operations at scale. They will likely coexist with the the foundational prerequisites.
more conventional ML- or DL-based AI applications,
which are better suited for analytical tasks such as Recognizing the untapped potential of AI, as well
anomaly detection, production analytics, setpoint as the associated challenges,10 and considering
optimization and forecasting. the pace of technological progress in this field, the
World Economic Forum, in collaboration with BCG,
AI-related advanced manufacturing technologies consulted with a community of industrial operations
and developments will continue to emerge, and technology executives, as well as experts
accompanied by new and even more diverse from academia, to co-develop a guidebook for
requirements – often at a faster pace than most harnessing the AI revolution in industrial operations.
companies can adopt them. One example of these The guidebook presented in this paper outlines
new innovations is quantum computing technology, a comprehensive approach for organizations to
which, like AI, will also have a disruptive impact on successfully navigate their AI transformation journey
industrial operations and on AI itself. in a fast-developing environment.

BOX 1. How different AI technologies relate to each other

– Machine Learning (ML): ML is a distinct subset of AI that allows systems to automatically learn and
improve from experience – data – without being explicitly programmed on specific rules; for example,
to predict values, calculate probabilities, identify cluster/groups and identify correlations.

– Deep Learning (DL): DL is a subset of ML, based on neural networks, and is suited to making deep
connections within the data based on large amounts of data and performing multiple calculations for
its features for each layer.

– Generative AI (GenAI): As a subset of DL, generative AI produces high-quality new content in

multiple formats such as text, images, code, videos and music.

Source: IBM11

Harnessing the AI Revolution in Industrial Operations: A Guidebook 8

2 Harnessing the
AI revolution in
industrial operations:
a guidebook
A new guidebook to systematically
approach and master the AI journey at
scale in industrial operations.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 9

 The implementation and, particularly, the successful to help manufacturers alleviate these challenges
scaling of AI in industrial operations is often by shedding light on best-practice approaches
hindered by a wide range of challenges. The adopted by leading manufacturers and technology
guidebook presented in this white paper aims experts on their AI journey (Figure 3).

FIGURE 3: Harnessing the AI revolution in industrial operations: a guidebook

Paving the way for success Mastering the AI journey Staying at the forefront of
from AI in industrial operations across production networks AI innovations

Value-adding AI applications for industrial operations

Foundations for AI implementation in industrial operations at scale

Organizational foundations Technological foundations

Governance Ecosystem partnering Data sourcing Computing

Skills and capabilities Legal compliance Data processing Connectivity

Change management and Applications and

Cybersecurity
communication user interfaces

Source: World Economic Forum in collaboration with BCG

The guidebook has five sections. The first three The final two sections comprise the building
sections represent the overarching stages of a blocks needed for a successful AI implementation
company’s AI journey: and scaling:

– Paving the way for success from AI in – Value-adding AI applications for industrial
industrial operations: Highlighting the operations: Pinpointing AI-based applications
importance of articulating an organization’s that address inefficiencies and operational
long-term AI objectives and transformation opportunities to achieve intended improvements
principles as a starting point on the AI journey. and outcomes.

– Mastering the AI journey across production – Foundations for AI implementation in

networks: Describing a step-by-step approach industrial operations at scale: Defining
from the status quo assessment to the design, the AI-related organizational and technology
engineering, implementation and scaling of foundations to achieve implementation at scale
value-adding AI applications and required and long-lasting success.
foundations.
While a white paper published in December
– Staying at the forefront of AI innovations: 202212 considered the various value-adding
Explaining the importance of conducting AI applications, this paper focuses on the
periodic AI reviews to continuously identify three stages of an AI journey and on the final
and integrate AI innovations and related stage, the AI-related organizational and
opportunities as they emerge. technological foundations.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 10

2.1 Paving the way for success from AI in
industrial operations

A successful AI journey starts by developing a depending on company-specific considerations,

clear understanding of the reasons for adopting some consistently come up as being a vital starting
AI and defining the vision and objectives of the point for success. An important principle to start
organization. Whether the vision is to increase with is the adoption of a value-driven approach
efficiency and agility, augment the capability of the with a consistent end-to-end perspective, to define
workforce or reduce emissions, grasping the “why” the AI target picture and subsequently identify the
streamlines the entire industrial transformation. related applications. To ensure a smooth rollout
Developing an AI strategy that allows the company across plants, it is also essential to think about
to achieve its targets aids in pinpointing the right AI the scalability of applications from the outset.
applications, facilitates transparent communication Additionally, to drive buy-in at various levels,
with all stakeholders and prevents misguided taking a pilot approach with rapid implementation
investment decisions. of use cases can demonstrate impact and value.
This test-and-learn approach can also highlight
In this context, aligning on the organization’s any adjustments needed for a successful rollout.
transformation principles can ensure that all Finally, keeping the organizational and technological
subsequent actions are directed towards the initial foundations in mind along the entire journey is
vision and objectives. While such principles vary essential to enable the use of AI at scale.

When Siemens introduced an AI-based predictive quality solution to increase surface-mount technology
(SMT) production, a value-driven approach with a clearly defined problem statement (describing the as-
is) and target definition (aligning the to-be) was followed. Having an aligned view on the status quo and
the objectives upfront helped over the course of the entire implementation, achieving a substantial lead
time reduction and quality improvement, as well as reducing energy consumption as a consequence of
an enhanced testing process. After the initial impact and with a higher level of process understanding,
Siemens has scaled the solution to additional plants and used the deployed infrastructure on other
products and machines to maximize the value generated.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 11

 2.2 Mastering the AI journey across
production networks

Unlocking the full potential of AI applications and change management and communication
requires going beyond initial pilots and deploying approach, as well as making design choices
solutions across the entire manufacturing along the technology infrastructure dimensions.
network. Based on the insights collected from
senior executives in operations and technology – Engineering: designing solutions and
experts, here are five key steps for identifying developing measures to close organizational
and implementing value-adding AI applications at and technology gaps – including assessing
scale and setting up their associated foundational potential solution vendors, engineering technical
enablers (Figure 4): solutions and proof of concepts and developing
a capability-building programme and technology
– Diagnosis: analysing the status quo and infrastructure solutions.
identifying gaps and opportunities – including
identifying and quantifying inefficiencies – Implementation: executing solutions and
and potential opportunities, understanding measures and validating their impact –
the required skills and capabilities and including implementing and validating proofs of
corresponding capability mismatches and concept for AI solutions and related technology
assessing the required technology prerequisites infrastructure in pilot areas and launching and
and the existing needs for action based on the testing the capability-building programme,
current technology infrastructure. governance and change management approach
in pilot plants.
– Design: defining the AI target picture,
strategy and roadmap – including determining – Scaling: rolling out validated AI solutions,
the most beneficial AI applications based on technology infrastructure and organizational
impact and required investment, deciding on measures – including monitoring of progress
a make-or-buy strategy and defining the AI- according to the roadmap across the entire
related governance, capability-building strategy production network.

FIGURE 4: Step-by-step approach for mastering the AI journey across production networks

1 2 3 4 5

Diagnosis Design Engineering Implementation Scaling

Analysing the status Defining the AI target Designing solutions Executing solutions Rolling out validated
quo and identifying picture, strategy and and developing and measures and AI solutions,
gaps and roadmap measures to close validating their technology
opportunities organizational and impact infrastructure and
technology gaps organizational
measures

Source: World Economic Forum in collaboration with BCG

Harnessing the AI Revolution in Industrial Operations: A Guidebook 12

2.3 Staying at the forefront of AI innovations

The rapid advancements in AI indicate that the business strategy and organizational
implementing and scaling AI will require more than priorities, manufacturers can continuously tap
a one-time effort. Recent breakthroughs – for into fresh opportunities and capture the full
example, innovations in generative AI – and the potential of AI. This also requires equipping the
new potential they unlock in manufacturing and organization with the right mechanisms and
supply chains highlight the importance of viewing resources to scout for the latest AI innovations
AI implementation as an ongoing journey. Industrial and opportunities affecting manufacturing and
organizations that fail to do so risk missing out on, supply chains. This process can be enabled by,
or belatedly adopting, new developments. for instance, an organizational structure that
includes a central team dedicated to supporting
Defining and implementing a systematic the entire AI journey across the production
approach for periodic AI reviews and maturity network. One of the mandates of this team
checks lays the foundation for staying at the would be to regularly review new developments
forefront of AI innovations, promoting long-lasting in AI – for example, by joining relevant
results and impact. By regularly verifying and platforms or attending corresponding events
updating the AI target picture and aligning it with and technology conventions.

2.4 Foundations for AI implementation in industrial

operations at scale

Strong organizational and technological This section provides an overview of these

foundations are essential for implementing and foundations and associated considerations derived
scaling AI applications across a production network from the leading organizations that contributed to
and achieving the goals of enhanced productivity, this guidebook’s development.
sustainability, flexibility and workforce support.

Organizational foundations

Drawing from their experiences with AI implementation and rollout of AI applications

implementation, experts emphasized that, across plants, while individual plants spearhead
while mechanisms to identify the most value- the actual implementation and are fully
adding applications are widely available, the accountable for their progress.
most significant challenges often pertain to the
organizational set-up itself. Within this context, – Roles and responsibilities: Using the RACI
five key AI-related dimensions were derived logic – which defines responsible, accountable,
from the consultations conducted: governance; consulted and informed stakeholders for tasks
skills and capabilities; change management and activities related to the AI journey – to
and communication; ecosystem partnering; support ownership and ensure execution in
and legal compliance. accordance with the roadmap. In addition
to considering the multiple roles needed, a
Governance common practice in industrial operations is to
designate initiative leads from specific plants to
Successful AI implementation and scaling require oversee AI applications and their cross-network
clear governance that is relevant to the efforts being implementation. These leads are supported by
undertaken. This governance builds the foundation an initiative coordinator from a central team.
through an adequate organizational structure,
clearly defined roles and responsibilities, required – Operational processes: Establishing clearly
processes and appropriate incentives and key defined processes, such as regular team
performance indicators (KPIs): meetings to monitor the implementation progress
or machine learning operations (MLOps)
– Organizational structure: Tailoring the procedures. In addition, depending on the AI
organizational structure to the company’s applications being implemented, it is essential
specific needs to support the AI journey. to review and adjust operational processes and
A common practice is to use a hybrid standard operating procedures to accommodate
approach: a central digital unit coordinates the the new requirements and workflows.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 13

– Incentives and KPIs: Defining targets based Given the nature of AI, its applications also require
on the overarching AI objectives and roadmap, development and implementation by cross-
and deriving suitable KPIs to monitor the AI functional teams with diverse expertise at the
implementation process and its impact. In convergence of IT, OT, data and AI technologies.
addition, it has proven beneficial to establish In addition to the primary technology-related
incentives for all stakeholders to drive ownership capabilities needed to develop, implement and
at all levels. Because AI initiatives tend to run operate AI applications in an industrial setting,
concurrently with routine operations, employees a set of soft skills has also proven to be
often need to allocate dedicated time to engaging instrumental in the transformation. These
in AI tasks and achieving defined targets. include eagerness to learn, problem-solving
capabilities and adaptability.
Skills and capabilities
Change management and communication
Having access to the right AI-related skills and
capabilities is instrumental for a successful AI The application of AI in the industrial context
journey and its sustainable scaling and impact. often encounters resistance, stemming from
Key capabilities and proficiencies that have scepticism caused by past failed efforts and the
emerged are: fear of potential job losses. Thus, proactive change
management and transparent communication from
– Data analysis: Executing data analysis, the outset of the AI journey are instrumental to
interpretation, visualization and design of ensuring success and sustaining long-term results.
training data for AI models and model validation.
Executives and experts contributing to this
– Data architecture: Defining AI-related data guidebook highlighted the below enablers
needs and the required data architecture to for successful change management and
facilitate data availability. communication:

– Data engineering: Realizing data availability, – Leadership buy-in: Leaders are aligned on the
usability and quality by building the requisite AI objectives, target picture, implementation
AI-related information technology (IT) and roadmap and necessary changes, and they
operational technology (OT), a data platform support the transformation.
and data pipelines to automate high-volume and
real-time data delivery. – Stakeholder engagement: Stakeholders at all
organizational levels have a clear understanding
– Data sciences: Automating complex data of their individual roles and their corresponding
analysis processes through the design and contributions to the success of AI.
development of statistical algorithms and
models, such as those based on ML. – Change culture: An environment that supports
and encourages the willingness to achieve
– Software engineering and development: the desired future state across all levels, with
Developing front-end and back-end software for leaders promoting and role-modelling the
AI applications and AI-based tools. desired behaviours.

– ML engineering: Bringing ML models to – Transparency and accessibility: AI applications,

production, which includes testing and including tailored stakeholder information
optimizing models, as well as establishing the materials, are available for all employees, thereby
MLOps infrastructure. enhancing the acceptability of AI.

– Automation engineering: Automating – Communication strategy: Authentic

shopfloor activities by implementing and communication, spearheaded by leaders
connecting industrial control hardware and throughout the organization and encouraging
using simulation tools for rapid prototyping a two-way dialogue that resonates with
of automation systems. employees on a personal level.

Conscious of the importance of change management and employee buy-in for the success of a new AI
solution, Körber Digital involved the affected employees right from the beginning of the adoption of the
FactoryPal solution and embedded it in their daily shopfloor interactions. This has been a key success
factor in the transformation, with the solution being scaled to more than 75 production lines and an
overall equipment efficiency improvement of 30%.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 14

Ecosystem partnering existing capability gaps, such as by
developing best-in-class data-processing
Considering the rapid evolution of AI, its inherent capabilities on modern digital infrastructure
complexity, the depth of expertise needed and or upskilling current talent.
the scarcity of available experts, collaborating with
different partners in the ecosystem and across value Legal compliance
chains has proven to be a key enabler for success.
This is especially true when contrasted with the The regulatory landscape is evolving and becoming
daunting alternative of managing every aspect of increasingly mature (e.g. the EU AI Act),13
the journey internally from the outset. Partnerships strengthening the importance of ensuring legal
spanning the entire industrial and AI ecosystem can compliance when implementing AI applications
be instrumental in, for example: in industrial operations. With legal responsibilities
varying across regions, the most effective
– Driving innovation: Leveraging industry best approaches involve identifying relevant compliance
practices and outside-in perspectives to provide markets and related policies and standards based
fresh insights that are valuable for defining an on the scope and scale of applications implemented
AI vision and corresponding target picture with across the production network. Depending on the
high-value applications, among other benefits. evaluation conducted and the risk level identified,
registration with oversight regulatory bodies
– Piloting: Using digital factories as testbeds as might be required. Additionally, it is important to
they offer AI immersion sessions and opportunities maintain up-to-date technical documentation in
to design, test and implement potential AI order to showcase compliance efforts and provide
applications in real factory environments. comprehensive details, ranging from the system
architecture to the risk management strategies
– Capability building: Collaborating with in place.
technology experts and academia to bridge

Technological foundations

Considering the digital nature of AI and the – Transactional layer: Encompassing the
importance of data to its various applications, an enterprise resource planning (ERP) system and
adequate existing or new technology infrastructure its warehouse management system (WMS),
needs to be in place to cater to AI requirements. transport management system (TMS) and PLM,
Based on the consultations conducted, six among others.
dimensions are most relevant: data sourcing;
data processing; applications and user interfaces; Data processing
computing; connectivity; and cybersecurity.
The first three dimensions are hierarchically The data processing infrastructure makes the data
interconnected, with data sourcing as the foundation, sources available for analytics and AI applications.
data processing in the intermediary position and The infrastructure encompasses:
applications and user interfaces as the top layer. The
latter three dimensions – computing, connectivity – Ingestion: Collecting and importing data from
and cybersecurity – run transversely to the first three. various sources, to be stored in a centralized
database. Techniques such as edge analytics or
Data sourcing IoT hubs are used to make (raw) data available
for subsequent processing.
AI applications are intrinsically reliant on data,
making it crucial for manufacturers to ensure they – Processing: Preparing and making data
have the necessary data sources readily available suitable for ML/DL algorithms and models
and accessible. These sources deliver different – e.g. through data cleaning, filtering and
types of data – e.g. structured data such as part contextualization.
descriptions in product life-cycle management (PLM),
or unstructured data such as time series data from – Storage: Storing (raw) data in a way that
sensors – and can typically be categorized into: creates a single source of truth for all
applications; for example, in a central data lake.
– Field layer: Providing data from programmable
logic controllers and internet of things (IoT) – Delivery: Constructing data pipelines to
devices and sensors, which can include, facilitate data flow and ensure its delivery to the
for example, real-time measurements of point of use, such as specific AI models.
temperature, vibrations or pressures.
– Archival and deletion: Using data repositories
– Control layer: Composed of the manufacturing to archive data and applying data governance
execution system (MES) and supervisory control for determinations about the ultimate deletion
and data acquisition (SCADA) systems. of data.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 15

Applications and user interfaces practices, including the model development,
testing, validation, deployment and servicing.
Another critical component of an AI-centric
technological infrastructure in manufacturing – Engagement layer: Consisting of both general
pertains to the AI tools and applications purpose and specialized applications, including
themselves, including: interfaces that allow AI application users to
interact with outcomes and insights in the
– Intelligence layer: Incorporating the algorithms manufacturing environment.
and AI models, which adhere to the MLOps

At Martur Fompak International, one important success factor in the implementation of an AI-based
quality control system – which reduced the quality control process from 58 to 2–3 seconds – was to
jointly develop with the employees a user interface tailored to the end user, which led to an increased
usability and acceptance by the workforce.

Computing Connectivity

AI applications, including the foundational ML and DL algorithms often rely heavily on (near)
analytical models such as ML and DL, often real-time communication and data availability for
work with vast amounts of data, necessitating many applications. As such, seamless networking
significant computing power to ensure timely and connectivity are essential, with both high
data processing. The speed at which these AI bandwidth and low latency to ensure scalability.
applications operate is largely determined by This can be achieved by establishing connectivity
the processors used, primarily central processing between plants via cloud links, within plants through
units (CPUs) and graphics processing units wired or wireless communication and between
(GPUs). There are two primary approaches to assets using appropriate IoT protocols.
providing AI with computing power: on-premises
and cloud-based. Cybersecurity

While the previously mentioned field and control AI processes a vast amount of data, some of
layers almost always operate on-premises, which can be sensitive. Ensuring data protection
numerous providers offer cloud solutions to throughout the entire process necessitates a
handle parts of the data processing and execute comprehensive cybersecurity strategy consisting of:
the AI applications. With the emergence of
generative AI solutions and the subsequent – Identity and access management (e.g. through
need to fine-tune foundational models using authentication management)
specific company data, several configurations
become viable. For instance, an open-source – Data protection, detection and response (e.g.
foundational model might be employed, using firewalls and intrusion detection systems)
fine-tuned and operated within a private
environment on-premises. Alternatively, a – Risk analysis and management
dedicated tenant within a provider’s cloud
platform could be used. – Recovery planning

Harnessing the AI Revolution in Industrial Operations: A Guidebook 16

Conclusion
The rapid evolution of AI is consistently yielding value-adding AI applications and required
important innovations – such as the recent foundations.
emergence of generative AI – offering tremendous
potential to enhance productivity and production – Staying at the forefront of AI innovations:
flexibility, drive sustainability and empower the Explaining the importance of conducting
workforce. However, while many companies have periodic AI reviews to continuously identify
begun integrating AI into their operations, most and integrate AI innovations and related
have not realized their AI-related targets or fully opportunities.
unlocked its full value.
Additionally, the guidebook elaborates on
With many manufacturers facing challenges when the enabling components that need to be at
implementing and scaling AI in their factories and the foundation of the AI journey from both
supply chains, this white paper’s guidebook aims organizational and technological perspectives,
to provide a systematic approach to harnessing the illustrated by select examples.
AI revolution in industrial operations. It describes
the primary focus of each stage of the journey and Moving forward, the World Economic Forum will
presents a systematic approach to designing the continue to shed light on the latest AI opportunities
AI target picture and implementing and scaling the and innovations affecting manufacturing and
associated AI applications and foundations: supply chains, as well as providing a unique
space for collaborations among industry leaders,
– Paving the way for success from AI in technology experts and academics to deploy AI
industrial operations: Highlighting the at scale in industrial operations, unlocking value
importance of articulating an organization’s for companies, society and the environment.
long-term AI objectives and transformation Stakeholders are invited to join this initiative to
principles as a starting point of the AI journey. share methodologies, incubate new pilots and
disseminate insights. The purpose of this endeavour
– Mastering the AI journey across production is to enable companies to successfully capture
networks: Describing a step-by-step approach the full value of AI in industrial operations and to
from the status quo assessment to the design, harness its ongoing innovations – today and in
engineering, implementation and scaling of the future.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 17

Contributors
World Economic Forum Boston Consulting Group

Maria Basso Daniel Küpper

Centre Curator, Advanced Manufacturing and Managing Director and Senior Partner, Global
Supply Chains (Co-)Lead BCG Manufacturing and Supply Chain,
Boston Consulting Group
Memia Fendri
Initiatives and Community Lead, Advanced
Manufacturing and Supply Chains

Kai-Frederic Seitz
Principal, Boston Consulting Group, seconded to
the World Economic Forum

Kyriakos Triantafyllidis
Head of Growth and Strategy, Centre for Advanced
Manufacturing and Supply Chains

Acknowledgements

The World Economic Forum thanks the following individuals for their contributions and participation in
working groups, interviews and community discussions:

Edward Atkins Jürgen Gotepass

Strategic Alliances, Tulip Interfaces, UK Chief Strategy Officer Manufacturing, European
Standardization & Industry Development
Gunter Beitinger Department, Huawei, Germany
Senior Vice-President Manufacturing, Head of
Factory Digitalization, Siemens, Germany Andreas Hauser
Chief Executive Officer, Digital Service, TÜV SÜD,
Frank L. Blaimberger Singapore
Vice-President, Global Head of Advanced
Manufacturing, TÜV SÜD, Germany Hagsoo Kim
Business Development Manager, MakinaRocks,
Andreas Böhmert South Korea
Partner, Knife Capital, South Africa
Dimitris Kiritsis
Aref Boualwan Professor Emeritus of ICT for Sustainable
Chief Initiatives and Startups Officer, Consolidated Manufacturing, Ecole Polytechnique Fédérale de
Contractors International Company, Greece Lausanne, Switzerland

Michelangelo Canzoneri Jay Lee

Global Head, Group Smart Manufacturing, Merck Clark Distinguished Professor and Director of
Group, Germany Industrial AI Center, Department of Mechanical
Engineering, University of Maryland, USA
Efe Erdem
Director, MESS Technology Platform, Turkish Natan Linder
Employers’ Association of Metal Industries (MESS), Chief Executive Officer, Tulip Interfaces, USA
Türkiye
Humera Malik
Marco Freidl Chief Executive Officer, Canvass AI, Canada
Head of Group Strategy and Business
Development, Oerlikon, Switzerland

Harnessing the AI Revolution in Industrial Operations: A Guidebook 18

Darko Matovski Rajiv Puri
Chief Executive Officer, causaLens, UK Vice-President, Manufacturing Strategy, Solutions
and Partnerships, Infosys, USA
Torbjørn Netland
Professor and Head of Chair of Production and Shwetha Shetty
Operations Management, ETH Zurich, Switzerland Product Owner, SAP Digital Supply Chain Solutions,
SAP, USA
Meirav Oren
Chief Executive Officer, Versatile, USA Daniel Szabo
Chief Executive Officer, Koerber Digital, Germany
Cyril Perducat
Senior Vice-President and Chief Technology Officer, Amogh Umbarkar
Rockwell Automation, USA Vice-President, SAP Product Engineering, SAP,
USA
Joris Poort,
Chief Executive Officer, Rescale, USA Cong Yu
Vice-President of Engineering, Celonis, USA

The World Economic Forum would also like to thank the following individuals for their collaboration and
support throughout the project:

Tunc Acarkan Harleen Kaur Sidhu

Technology Governance Director, Turkish AI Research Engineer, Digital Service, TÜV SÜD,
Employers’ Association of Metal Industries (MESS), Singapore; Project Fellow seconded to the World
Türkiye Economic Forum

Ece Akin Armutak

Technology Governance Manager, Turkish
Employers’ Association of Metal Industries (MESS),
Türkiye

Harnessing the AI Revolution in Industrial Operations: A Guidebook 19

 Endnotes

1 World Economic Forum, 4 Ways Artificial Intelligence Could Transform Manufacturing, January 2023:
https://www.weforum.org/agenda/2023/01/4-ways-artificial-intelligence-manufacturing-davos2023/.
2 Boston Consulting Group, Global Survey on AI in Industrial Operations 2023 (N=1,800), April 2023:
https://www.bcg.com/about/partner-ecosystem/world-economic-forum/ai-project-survey.
3 World Economic Forum, Beyond the Status Quo: How Generative AI Will Transform Industrial Operations, June 2023:
https://www.weforum.org/agenda/2023/06/how-generative-ai-will-transform-industrial-operations.
4 World Economic Forum, 4 Ways Artificial Intelligence Could Transform Manufacturing, January 2023:
https://www.weforum.org/agenda/2023/01/4-ways-artificial-intelligence-manufacturing-davos2023/.
5 World Economic Forum, Unlocking Value from Artificial Intelligence in Manufacturing, December 2022:
https://www.weforum.org/whitepapers/unlocking-value-from-artificial-intelligence-in-manufacturing/.
6 Boston Consulting Group, Global Survey on AI in Industrial Operations 2023 (N=1,800), April 2023:
https://www.bcg.com/about/partner-ecosystem/world-economic-forum/ai-project-survey.
7 Ibid.
8 NVIDIA, NVIDIA Hopper, Ampere GPUs Sweep Benchmarks in AI Training, November 2022:
https://blogs.nvidia.com/blog/2022/11/09/mlperf-ai-training-hpc-hopper/.
9 SAP, Artificial Intelligence: https://www.sap.com/products/artificial-intelligence/supply-chain.html?pdf-asset=7ac887f0-
4d7e-0010-bca6-c68f7e60039b&page=1&source=social-atw-mailto.
10 World Economic Forum, Unlocking Value from Artificial Intelligence in Manufacturing, December 2022:
https://www.weforum.org/whitepapers/unlocking-value-from-artificial-intelligence-in-manufacturing/.
11 IBM, Machine Learning (accessed September 2023): https://www.ibm.com/design/ai/basics/ml/#:~:text=Machine%20
learning%20(ML)%20is%20a,teaching%20machines%20how%20to%20learn; IBM Research, What is Generative AI?,
April 2023: https://research.ibm.com/blog/what-is-generative-AI.
12 World Economic Forum, Unlocking Value from Artificial Intelligence in Manufacturing, December 2022:
https://www.weforum.org/whitepapers/unlocking-value-from-artificial-intelligence-in-manufacturing/
13 European Parliament, EU AI Act: First Regulation on Artificial Intelligence, June 2023: https://www.europarl.europa.eu/
news/en/headlines/society/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence.

Harnessing the AI Revolution in Industrial Operations: A Guidebook 20

The World Economic Forum,
committed to improving
the state of the world, is the
International Organization for
Public-Private Cooperation.

The Forum engages the

foremost political, business
and other leaders of society
to shape global, regional
and industry agendas.

World Economic Forum

91–93 route de la Capite
CH-1223 Cologny/Geneva
Switzerland
Tel.: +41 (0) 22 869 1212
Fax: +41 (0) 22 786 2744
contact@weforum.org
www.weforum.org