Fostering Human-AI Collaboration with Digital Intelligent Assistance in Manufacturing SMEs

Abstract

Greater cognitive task load and the growing shortage of highly skilled labor provide ground for assistance systems based on Artificial Intelligence (AI). Conventional graphical interfaces to such systems are often hard to understand, obtrusive, and unintuitive. Natural language interactions provide one approach to address this shortcoming. Recently, voice-enabled Digital Intelligent Assistants (DIAs) for manufacturing matured enough to satisfy various industrial requirements. Their adoption by SMEs, however, is challenging due to the high cost of developing, deploying, and maintaining them. This paper presents the vision of a white-label shop for DIAs and human-AI collaboration in manufacturing. This shop and its associated concepts seek to reduce costs by introducing a one-stop shop where SMEs can find various elements necessary to introduce DIAs in their organizations.

Full text

Fostering Human-AI Collaboration with Digital
Intelligent Assistance in Manufacturing SMEs
Stefan Wellsandt1, Mina Foosherian1, Alexandros Bousdekis2, Bernhard Lutzer3, Fotis
Paraskevopoulos2, Yiannis Verginadis2,4, and Gregoris Mentzas2
1BIBA - Bremer Institut für Produktion und Logistik GmbH at the University of Bremen, Bre-
men, Germany
2Information Management Unit (IMU), Institute of Communication and Computer Systems
(ICCS), National Technical University of Athens (NTUA), Athens, Greece
3TTTech Industrial Automation AG, Vienna, Austria
4Athens University of Economics and Business, Athens, Greece
Abstract. Greater cognitive task load and the growing shortage of highly skilled
labor provide ground for assistance systems based on Artificial Intelligence (AI).
Conventional graphical interfaces to such systems are often hard to understand,
obtrusive, and unintuitive. Natural language interactions provide one approach to
address this shortcoming. Recently, voice-enabled Digital Intelligent Assistants
(DIAs) for manufacturing matured enough to satisfy various industrial require-
ments. Their adoption by SMEs, however, is challenging due to the high cost of
developing, deploying, and maintaining them. This paper presents the vision of
a white-label shop for DIAs and human-AI collaboration in manufacturing. This
shop and its associated concepts seek to reduce costs by introducing a one-stop
shop where SMEs can find various elements necessary to introduce DIAs in their
organizations.
Keywords: voice assistant, artificial intelligence, augmented intelligence, hu-
man-AI collaboration, manufacturing SMEs.
1 Introduction
Manufacturers in Europe face increasingly volatile market environments shaken by re-
gional and global political, economic, and social crises. These crises result from natural
disasters, extreme weather intensified by climate change, oppressive regimes, and war.
Organizations and society would benefit if resilience mechanisms to recover from such
crises were available. Under the circumstances above, Small and Medium-sized Enter-
prises (SMEs) and mid-caps in manufacturing face extraordinary challenges. They of-
ten produce components for more complex products or operate in niche or emerging
markets for special-purpose machinery and equipment. Both characteristics drive
SMEs and mid-caps to be exceptionally careful in spending their resources on activities
with uncertain costs or outcomes. Operating in niche markets also makes it difficult for
them to hire skilled workers when necessary. Lower salaries compared to large

2
enterprises and a generally smaller available workforce caused by demographic change
in Europe are vital concerns.
Greater cognitive task load and growing shortage of high-skilled labor call for new
smart interactions between the cyber-physical production system and the Operator 4.0
[1,2] since conventional modes of interaction make this difficult since they are often
hard to understand, obtrusive and unintuitive [3]. Therefore, new interactions between
operators and machines are needed in the hybrid-augmented intelligence paradigm
[4,5]. This paradigm can be realized through voice-enabled Digital Intelligent Assis-
tants (DIAs), providing fast, intuitive, and potentially hands-free access to systems
through voice-based interaction and cognitive assistance [4-6]. Although industrial ap-
plications of DIAs have emerged only recently, they are expected to play a significant
role in the collaboration between humans and AI systems [1,8,9]. Apart from the exist-
ing technological challenges, this direction requires new approaches and business mod-
els for the expansion of such technological solutions to facilitate the adoption by a
larger amount of industries, particularly SMEs.
In this paper, we present the vision of the WASABI (White-label shop for digital
intelligent assistance and human-AI collaboration in manufacturing) project, which fo-
cuses on intelligent digital assistance solutions in order to help humans achieve their
goals without marginalizing them - this will contribute to human-centered manufactur-
ing. The scope of this assistance is on augmenting the workers’ cognitive capabilities
with decision-support systems and accelerating knowledge acquisition and transfer
through Natural Language Processing (NLP). Such assistance can lead to hybrid-aug-
mented intelligence [4], where humans and software contribute their intelligence for
mutual benefits. WASABI aims to provide SMEs with the tools and knowledge to im-
prove workers’ capacities and performance, providing advanced user interfaces for con-
tinuous augmented-hybrid decision-making. A network of Digital Innovation Hubs
(DIHs) that will help boost impact by guiding SMEs in this new path will be created
and integrated within other DIH networks. Our customized, federated, white-label shop
will include DIAs and skill packages to help organizations reach their sustainability
goals. Blue-collar and white-collar workers will be capable of using it for hands-free or
eyes-free computer interaction, AI-based advice and guidance, and augmented analyt-
ics.
The rest of the paper is organized as follows. Section 2 provides a literature review
on human-AI collaboration with DIAs in manufacturing. Section 3 presents the vision
of the WASABI project. Section 4 presents the main technological pillars on which the
WASABI solution will be built. Section 5 demonstrates the use cases in which the so-
lution will be evaluated. Section 6 concludes the paper and outlines future work.
2 Human-AI collaboration with Digital Intelligent Assistants
During the last years, there is a growing interest in the development of voice-enabled
assistants for the manufacturing environment in order to tackle with the distinct chal-
lenges compared to assistants addressing consumers (e.g., Amazon Alexa or Google
Assistant). In [9], the research work presented expectations, requirements, and a

3
concept for a voice-enabled DIA that supports maintenance activities. In [10], the re-
search work presented an overview of assistants’ benefits in manufacturing, including
central access, customization, delegation and guidance, eyes-free and hands-free inter-
actions, mobile assistance, the support of multiple interface types, permanent accessi-
bility, and speed. Moreover, they outline the factors that determine the use of a digital
assistant, including trust in AI systems, impacts on teams and individuals, training and
education, and capabilities of open and closed technologies. In [2], the landscape of
digital assistants in terms of human cyber–physical systems is analyzed, with the aim
of proposing a conceptual framework at the cognitive level. In [7], a framework for the
evaluation of voice-enabled AI solutions in Industry 5.0 is proposed. It consists of four
dimensions: the trustworthiness of the AI system; the usability of the DIA; the cognitive
workload of individual users; and the overall business benefits for the corporation. In
[3], an approach is proposed in order to enable maintenance experts and operators to
interact with a predictive maintenance system through a DIA, while, in [4], the same
authors extended their work in the context of hybrid-augmented intelligence to integrate
human knowledge into the predictive maintenance process.
In [11], the research work investigated how software robots, also known as softbots,
can support the Operator 4.0 in smart factory environments, helping in the interfacing
between smart machines and computer information systems with the aims of supporting
the Operator 4.0 in different tasks at the shop floor. In [8], the use of Collaborative
Networks foundations at the intra-organizational level is outlined, applying them in the
support of collaborative softbots. They implemented five use cases and demonstrated
the potentials of better human-automation symbiosis when groups of CPSs, information
systems and humans have to cooperate and collaborate using collaborative softbots to
improve operational excellence and human satisfaction in smart, social factories. In [1],
the research work shed light on the prospective adoption and acceptance of voice assis-
tants to assist the Operator 4.0 during industrial production processes within the Social
Smart Factory. Leveraging on quality-driven engineering of human–machine interac-
tion systems and on the prototyping of a voice-enabled assistant for a CNC milling
machine, insights and challenges are discussed. In [12], the research work proposed a
softbot with chatting capabilities in order to enable managers to identify their main
problems, suggests supporting actions.
In [5], the authors presented a voice-enabled DIA, which interacts with a prescriptive
quality analytics service in the context of Augmented Manufacturing Analytics in order
to allow operators to access and customize defect predictions and the prescribed qual-
ity-related actions. In [13], the research work proposed an intelligent assistance system
enabling employees to help themselves in the workplace and provide them with com-
petence-related support. In [14], they developed an AI-powered pervasive system that
provides cognitive augmentation to users of complex systems. The paper presented an
AI cognitive assistant that provides on-the-job training to novices while acquiring and
sharing (tacit) knowledge from experts. Cognitive support is provided as dialectic rec-
ommendations for standard work instructions, decision-making, training material, and
knowledge acquisition. These recommendations are adjusted to the user and context to
minimize interruption and maximize relevance. In [15], the research work proposed
CLAICA, a Continuously Learning AI Cognitive Assistant that learns from

4
experienced workers, formalizes new knowledge, stores it in a knowledge base, along
with contextual information, and shares it when relevant. Their results provided a
deeper understanding of how prior training, context expertise, and interaction modality
affect the user experience of cognitive assistants. They also draw on their results to
elicit design and evaluation guidelines for cognitive assistants that support knowledge
exchange in fast-paced and demanding environments, such as an agile production line.
3 Vision and Rationale
Digital assistance and conversational AI will become standard practices to reach sus-
tainability and other goals in manufacturing. Humans will use them, for instance, to
identify and assess opportunities to turn waste into a resource and to reorganize work
to minimize carbon footprints. Access to these benefits will be as simple as selecting
and configuring Apps from an online store, and interoperability minimizes vendor lock-
in and maximizes information valorization. New AI-focused training services for em-
ployees will be a general practice too. They let workers experience solutions and teach
them about AI’s capabilities, risks, and limitations in manufacturing. Digital assistance
solutions will blend into Europe’s emerging legal framework for AI and will be afford-
able and manageable even for small producers.
3.1 Digital assistance purpose and maturity
Assistance will focus on augmenting the workers’ cognitive capabilities with decision-
support systems and accelerating knowledge acquisition and transfer through NLP
[4,16]. Augmentation will help workers identify, assess, and take the opportunities to
improve their organization’s performance indicators. Since sustainability increasingly
becomes a competitive advantage for producers, decision-support systems are widely
available for reuse, remanufacturing, and recycling [17]. Digital assistance will support
humans in using related decision-support systems, potentially creating hybrid-aug-
mented intelligence, where humans and software contribute their intelligence for mu-
tual benefits [18,19].
Knowledge acquisition and transfer support focuses on increasing the available work-
force by upskilling and augmenting people to integrate them faster into manufacturing
companies. In this context, upskilling is an acknowledged approach to building a resil-
ient economy [20], and so-called plug-and-play workers are an approach to increase
resilience in manufacturing and supply chains [21]. Such workers utilize digital tech-
nology, like DIAs, to minimize learning times. Eventually, assistance to acquire or im-
prove skills will help regions, supply chains, and individual organizations recover from
crises like the COVID-19 pandemic and war.
AI techniques are essential to the perceived intelligence of digital assistance solu-
tions. However, implementing them as trustworthy AI systems is a comparably new
requirement and as much a technical challenge as it is an organizational and legal one
[22]. Despite this promising outlook, clear success stories of digital assistance solutions
with conversational interfaces are sparse, especially in SMEs. Large software

5
companies, such as Google and Amazon, and open source initiatives, such as Mycroft
AI (https://mycroft.ai/), Mozilla’s Common Voice, and Coqui (https://coqui.ai), only
create the essential tools for these solutions. Their preparatory work concentrates on
the consumer sector, and only recently specialized initiatives such as COALA (COgni-
tive Assisted agile manufacturing for a LAbor force supported by trustworthy Artificial
Intelligence) (www.coala-h2020.eu) or innovative companies, such as SPIX Industry
(www.spix-industry.com), customize these tools for manufacturing.
The global market for such solutions is hard to determine because e-commerce and
customer service applications overshadow it. A market survey from 2023 does not list
manufacturing as a separate market vertical [23]. DIAs for manufacturing are, there-
fore, out of the lab but not ready to be widely adopted by SMEs and mid-caps. Reasons
identified during the COALA research project include complex and thus costly deploy-
ment procedures, skepticism about the cost-benefit ratio of human-AI collaboration,
and a lack of organizational measures to up-skill workers in managing capabilities,
risks, and limitations of AI-based assistance technology.
3.2 Assistant skills and building a federated distribution channel
DIAs rely on skills, i.e., software packages, to perform general and solve specific prob-
lems, e.g., in sustainable, agile, and resilient manufacturing. Skills for sustainability or
resilience are not readily available because researchers and IT companies began exper-
imenting with complete assistants in manufacturing only recently [2-6]. One barrier to
developing more practical skills is that experts from multiple domains, such as indus-
trial engineering, computer engineering, media informatics, didactics, education, and
AI law, must collaborate closely with manufacturers and their workers. This interdisci-
plinary approach is costly but necessary to ensure an adequate human-centered and ef-
ficient solution. Creating successful skills also requires continuously improving train-
ing data and dialog rules based on many actual conversations. This constraint exists
because it is hard to predict what users will say to an assistant in a new domain. Any
larger step forward in this area will require more experiments with actual manufacturing
companies.
Organizations create skills for a specific assistant framework, such as Mycroft, Siri,
Alexa, or Google Assistant. This approach creates a vendor lock-in where customers
must purchase skills per framework while skill developers must maintain multiple skill
versions. Both behaviors increase the cost of digital assistant skills and reduce the utility
of new assistant frameworks that start with very few skills. Besides, they increase the
existing frameworks’ market dominance until the lack of competition slows innovation
and increases prices. Typically, several skills are necessary to assist users effectively.
Providers of digital assistance solutions could offer these skill packages to customers
along with relevant hardware and a broad range of services, such as application poten-
tial analysis, employee training, deployment, maintenance, and customization. How-
ever, the lack of a distribution channel for such solutions slows the uptake of DIAs.
A white-label online shop to offer assistance solutions could overcome this barrier.
This shop would provide federation capabilities to facilitate growth and solution diver-
sity, a royalty-sharing mechanism for skills to increase skill exposure, and a

6
demonstrator for interoperable skills. The shop could base on an acknowledged open-
source shop framework, such as PrestaShop. It should offer everything necessary to
deploy and operate DIAs, including the software, hardware, and support services.
4 Technological Pillars
This Section describes the main technological pillars to realize the vision above. These
pillars are (i) digital assistance technology for manufacturing (Section 4.1); (ii) feder-
ated shop for industrial digital assistance solutions (Section 4.2); (iii) COALA assistant
(Section 4.3); (iv) Non-Fungible Token (NFT) module (Section 4.4); (v) edge compu-
ting solution (Section 4.5).
4.1 Digital assistance technology for manufacturing
An assistant supports or entirely takes over time-consuming, stressful, or otherwise un-
desirable activities for the client [9]. A digital assistant (DA) does the same through
software for its user(s). WASABI focuses on DAs, that are also conversational agents.
A conversational agent is an application that accepts user input in the form of voice or
text and provides responses in natural language. Assistants on mobile devices and
Smartspeakers with screens allow rich media responses. Besides, conversational de-
signers can build dialogs so that the assistant has one or more personas to interact with
various user groups of age, gender, and social, educational, and cultural backgrounds.
Fig. 1 illustrates the working principle of a digital assistant for the industry.
Fig. 1. Working principle of a voice-enabled DA for the industry.
4.2 Federated shop for industrial digital assistance solutions
The federated shop for industrial digital assistance bases on PrestaShop, a large and
acknowledged open-source project (https://github.com/PrestaShop). We will extend it
to create a distribution channel for digital assistance solutions. It will focus on digital
assistant technology, which includes skills, hardware, services, and utility software.

7
One skill can belong to several skill packages. Skills and packages will have licenses
defining the payment terms, including free use. All skills using AI models will contain
an explainability mechanism to increase trustworthiness. The shop will contain virtual
shelves filled with existing results from other research and innovation projects and com-
mercial tools and services. Fig. 2 illustrates the main shop elements, the extensions,
and the existing assistant’s base infrastructure.
Fig. 2. Federated white-label online shop for digital assistance solutions and base infrastructure.
4.3 COALA assistant
The EU-funded research project COALA developed a DIA for augmented data analyt-
ics [5] and cognitive support [15], along with a prototype for flexible AI-decision ex-
planations [24]. It supports factory workers and supervisors that need to use analytics
tools or train on-the-job. The assistant’s core is the privacy-focused, open-source voice
assistant Mycroft coupled with chatbots built with the open-source framework Rasa as
it is shown in Fig. 3.
The core assistant provides basic features needed for corporate assistants, such as
GDPR compliance, security measures, and fundamental intents for conversation con-
trol. It comes with a customized open-source Android app to connect to Mycroft. This
custom app provides QR code reading, configuration options, and three conversation
modes. Mycroft Core provides the overall privacy-focused framework (e.g., skills,
management tools, and message bus). Rasa provides powerful language processing
tools and patterns to build chatbots. The COALA assistant uses custom Rasa chatbots
for skills that require more complex language understanding and interactions. An iden-
tity management service secures the infrastructure, and a GraphQL server manages the
access to external data exchange interfaces.

8
Fig. 3. The high-level architecture of the COALA assistant.
4.4 NERVE platform and edge devices
WASABI will include an edge computing solution called NERVE platform
(https://www.tttech-industrial.com/nerve) to deploy and operate AI-based solutions at
the edge. It consists of three core elements: The node software running on several edge
device types, qualified devices, and the NERVE management system. These elements
allow data acquisition from the shop floor and subsequent data analysis. The node soft-
ware enables applications to run on devices through Docker containers, virtual ma-
chines, or a Soft-PLC and connects to the central management software. Edge devices
are typically industrial PCs qualified to run the node software. The management system
has three tasks. First, it provides remote access to the edge nodes for software life cycle
management and provisioning. Second, it collects data from the devices, provides re-
mote analysis, and fog computing services (seamless distribution of the computing
workload between edge and cloud) or pushes data to another consumer, such as a data-
base. Third, it deploys workloads to edge devices.
4.5 Non-Fungible Token (NFT) module
The Non-Fungible Token (NFT) module will use NFT-based smart contracts to manage
skill royalties. This is necessary because no central entity manages the payment of skills
in the federated shop ecosystem. The NFT module will create and assign NFTs to skills
and use them to distribute royalties to skill developers when shops sell a skill acquired
via federation. This module will interact with a public distributed ledger like Ethereum
or IOTA. NFTs and a public ledger provide decentralized skill traceability – no single
entity controls or fails to compensate skill providers. It aims at simplifying or even
incentivizing skill-sharing among shops. NFTs differ from cryptocurrencies because

9
their identification codes reflect additional information, including metadata describing
a digitized asset, i.e., skills in WASABI’s case.
Every time a skill is used, its developer (and its respective token) can get credit for
it in the form of the cryptocurrency supported by the blockchain network. The skill
developer can be rewarded by interacting with the contract and transferring the respec-
tive amount. In such a case, this amount is locked in the contract until the skill developer
transfers it to their wallet. Mapping skills to NFTs can streamline the WASABI shop
processes, eliminate intermediaries, and bolster security. Despite the benefits of NFTs,
they are in a relatively immature state; hence, the market for NFTs is not very liquid.
This means NFTs may be difficult to trade, especially during periods of distress. To
mitigate this risk, WASABI will also employ conventional means of payment/dona-
tions.
5 Use Cases
The vision above will be evaluated in the context of three use cases: augmented waste
management and valorization (Section 5.1), assisted workforce management after re-
gional and global crises (Section 5.2), and assisted quality assurance for sustainable
products (Section 5.3).
5.1 Use case 1: Augmented waste management and valorization
This use case focuses on waste management in manufacturing to valorize waste. If a
production process cannot avoid waste for technical reasons, this waste may be reusable
by other organizations, provided its characteristics meet their requirements. Advanced
user interfaces will help workers describe production waste to identify if it qualifies for
reuse by a third party. Workers will describe waste through a mobile conversational
interface and rich-media records, such as images and videos of waste and relevant ma-
chinery. If users require specific expert-level waste descriptions, e.g., a laboratory anal-
ysis, they will receive suggestions on contacting local experts. This use case focuses on
increasing sustainability by significantly simplifying waste valorization. WASABI will
demonstrate its solution by customizing and connecting the COALA assistant to the
rEUse waste management platform (http://www.fibereuse.eu/). This will be applied in
two cases: (i) in dimensional metrology, we will develop a module that measures man-
ufacturing scrap pieces to see if other manufacturing SMEs or midcaps can reuse them.
The AI assistant shows how the part differs from the aimed part (dimensional defects),
if it can be remanufactured for reuse, or if it needs to be used for other purposes. On the
rEUse platform, further information will be obtained on how to recycle this part. (ii) in
recycling and revamping surgical tools, the use case will check the instruments used
after each surgery following SOPs (Standardized Operation Procedures). If these tools
are still fit for use, they are sterilized (following another SOP) and used again, and if
they are not, the information on the tool is published on the rEUse platform for it to be
recycled.

10
5.2 Use case 2: Assisted workforce management after regional and global
crises
This use case focuses on manufacturing workforce management after a regional crisis
that forces many people to search for jobs (e.g., war, extreme weather, and natural dis-
asters). Such a crisis may cause people to flee or migrate to or within Europe. People
searching for jobs will likely possess very different job experiences, education, ethnic,
social, demographic backgrounds, and languages, making it challenging to integrate
them at work. Manufacturing organizations that suffer from workforce shortages (e.g.
due to demographic change or rapidly increasing demand) can use human-centered dig-
ital assistance solutions to onboard and integrate interested people faster into their
workforce. Key characteristics of these solutions are multilingual conversational inter-
faces, customizable personas, frustration-mitigation mechanisms, and adaptable assis-
tance based on learning progress. This use case aims to increase societal resilience after
a crisis and increase agility by rapidly increasing and upskilling an organization’s work-
force. A second effect is that employers could integrate new workers faster and free up
the time of the existing employees resulting in productivity gains. WASABI will
demonstrate its solution in this use case by customizing the COALA assistant’s dialog
model and creating a knowledge base (e.g., Neo4J graph database) that contains what
new workers should focus on during the onboarding process to enhance the integration
and participation of workers such as foreign employees. An important non-technical
challenge is maintaining the power balance between the new employees and their em-
ployers. This is relevant because the employer can control the DIA behavior without
the employee recognizing it.
5.3 Use case 3: Assisted quality assurance for sustainable products
This use case focuses on the augmentation of product quality testing to increase product
and worker safety, carbon footprint, and workers’ cognitive skills. The latter aims to
reduce the burden of repetitive, boring, error-prone, and knowledge-intensive activities.
A DIA will support workers in product testing by interactively executing a validation
protocol that assures the highest safety and sustainability standards while at the same
time reducing validation time and energy consumption. To this end, the assistant will
be able to learn the main product stress characteristics from past testing data and gen-
erate synthetic high-throughput product quality testing datasets. During the test defini-
tion phase, the assistant will help the operator select the dataset that best suits the spe-
cific testing conditions, European directives, and international safety, performance, and
sustainability standards. During the testing phase, operators will talk to the assistant to
receive helpful information about testing intermediate results. The assistant will collect
user feedback used for training data labeling. A continuous and informative feedback
loop, and an adaptive validation protocol, will allow the assistant and the operator to
collaborate to increase the sustainability of the product quality testing process.
WASABI will demonstrate its solution in four cases: (i) automotive battery testing in a
new facility for battery testing (several climatic chambers, with several cycling charg-
ing and discharging functionalities and various monitoring and safety features) already

11
equipped with the most advanced technologies for digital twin and safety standards; (ii)
dimensional metrology to increase the quality of using of a coordinate measuring ma-
chine and the measurement performed, and improving tool preparations and the meas-
urement process; (iii) personal protection equipment testing to assist workers during
the process and product quality checks; and, (iv) prosthetics quality testing to check the
prosthetics at the end of the production cycle and assistance during result reporting.
6 Conclusions and Future Work
This paper presents the vision of the WASABI project, which focuses on intelligent
digital assistance solutions to help humans achieve their goals without marginalizing
them - this will contribute to human-centered manufacturing. The WASABI solution
will be built upon five technological pillars: digital assistance technology for manufac-
turing, federated shop for industrial digital assistance solutions, COALA assistant, NFT
module, and edge computing solution. The solution will first be evaluated using aug-
mented waste management and valorization, assisted workforce management after re-
gional and global crises, and assisted quality assurance for sustainable products. Then,
in the context of the project, we will set up two rounds of open calls for up to 20 exper-
iments to test digital assistance solutions or contribute new ones with an emphasis on
supporting SMEs in adopting the technology.
Acknowledgments. This work is partly funded by: (i) the European Union’s Horizon
2020 project COALA “COgnitive Assisted agile manufacturing for a LAbor force sup-
ported by trustworthy Artificial Intelligence” (Grant agreement No 957296); and (ii)
the European Union’s Horizon Europe project WASABI “White-label shop for digital
intelligent assistance and human-AI collaboration in manufacturing” (Grant agreement
No 101092176). The work presented here reflects only the authors’ view, and the Eu-
ropean Commission is not responsible for any use that may be made of the information
it contains.
References
1. Longo, F., Padovano, A.: Voice-enabled Assistants of the Operator 4.0 in the Social Smart
Factory: Prospective role and challenges for an advanced human–machine interaction. Man-
ufacturing Letters, 26(1), 12-16 (2020).
2. Angulo, C., Chacón, A., Ponsa, P.: Towards a cognitive assistant supporting human opera-
tors in the Artificial Intelligence of Things. Internet of Things, 100673 (2022).
3. Wellsandt, S., Klein, K., Hribernik, K., Lewandowski, M., Bousdekis, A., Mentzas, G., Tho-
ben, K. D.: Towards Using Digital Intelligent Assistants to Put Humans in the Loop of Pre-
dictive Maintenance Systems. IFAC-PapersOnLine 54(1), 49–54 (2021).
4. Wellsandt, S., Klein, K., Hribernik, K., Lewandowski, M., Bousdekis, A., Mentzas, G., Tho-
ben, K. D.: Hybrid-augmented intelligence in predictive maintenance with digital intelligent
assistants. Annual Reviews in Control 53(1), 382-390 (2022).
5. Bousdekis, A., Wellsandt, S., Bosani, E., Lepenioti, K., Apostolou, D., Hribernik, K.,
Mentzas, G.: Human-AI Collaboration in Quality Control with Augmented Manufacturing

12
Analytics. In: Advances in Production Management Systems (APMS). Artificial Intelli-
gence for Sustainable and Resilient Production Systems. Springer International Publishing,
Cham, pp. 303–310 (2021).
6. Rabelo, R. J., Romero, D., Zambiasi, S. P., Magalhães, L. C.: When softbots meet digital
twins: Towards supporting the cognitive operator 4.0. In: Advances in Production Manage-
ment Systems (APMS). Artificial Intelligence for Sustainable and Resilient Production Sys-
tems, pp. 37-4. Springer International Publishing (2021).
7. Bousdekis, A., Mentzas, G., Apostolou, D., Wellsandt, S.: Evaluation of AI-Based Digital
Assistants in Smart Manufacturing. In Advances in Production Management Systems
(APMS) 2022. Proceedings, Part II (pp. 503-510). Cham: Springer Nature Switzerland
(2022).
8. Rabelo, R. J., Romero, D., Zambiasi, S. P.: Softbots Supporting the Operator 4.0 at Smart
Factory Environments. In Advances in Production Management Systems. Smart Manufac-
turing for Industry 4.0. Springer International Publishing, Cham, 456–464 (2018).
9. Wellsandt, S., Foosherian, M., and Thoben, K. D.: Interacting with a Digital Twin using
Amazon Alexa. Procedia Manufacturing 52, 4–8 (2020).
10. Wellsandt, S., Hribernik, K., Thoben, K. D.: Anatomy of a digital assistant. In Advances in
Production Management Systems. Artificial Intelligence for Sustainable and Resilient Pro-
duction Systems: IFIP WG 5.7 International Conference, APMS 2021, Nantes, France, Sep-
tember 5–9, 2021, Proceedings, Part IV (pp. 321-330). Springer International Publishing
(2021).
11. Rabelo, R. J., Zambiasi, S. P., Romero, D.: Collaborative softbots: enhancing operational
excellence in systems of cyber-physical systems. In Collaborative Networks and Digital
Transformation: 20th IFIP WG 5.5 Working Conference on Virtual Enterprises, PRO-VE
2019, Turin, Italy, September 23–25, 2019, Proceedings 20 (pp. 55-68). Springer Interna-
tional Publishing (2019).
12. Abner, B., Rabelo, R. J., Zambiasi, S. P., Romero, D.: Production management as-a-service:
a softbot approach. In Advances in Production Management Systems. Towards Smart and
Digital Manufacturing: IFIP WG 5.7 International Conference, APMS 2020, Novi Sad, Ser-
bia, August 30–September 3, 2020, Proceedings, Part II (pp. 19-30). Springer International
Publishing (2020).
13. Deppe, S., Brandt, L., Brünninghaus, M., Papenkordt, J., Heindorf, S., Tschirner-Vinke, G.:
AI-Based Assistance System for Manufacturing. In 2022 IEEE 27th International Confer-
ence on Emerging Technologies and Factory Automation (ETFA) (pp. 1-4). IEEE (2022).
14. Freire, S. K., Panicker, S. S., Ruiz-Arenas, S., Rusák, Z., Niforatos, E.: A Cognitive Assis-
tant for Operators: AI-Powered Knowledge Sharing on Complex Systems. IEEE Pervasive
Computing (2022).
15. Freire, S. K., Niforatos, E., Wang, C., Ruiz-Arenas, S., Foosherian, M., Wellsandt, S., Boz-
zon, A.: Lessons Learned from Designing and Evaluating CLAICA: A Continuously Learn-
ing AI Cognitive Assistant. In Proceedings of the 28th International Conference on Intelli-
gent User Interfaces (pp. 553-568) (2023).
16. Wellsandt, S., Foosherian, M., Lepenioti, K., Fikardos, M., Mentzas, G., Thoben, K. D.:
Supporting Data Analytics in Manufacturing with a Digital Assistant. In: Advances in Pro-
duction Management Systems (APMS). Artificial Intelligence for Sustainable and Resilient
Production Systems. Springer International Publishing, pp. 511-518 (2022).
17. Acerbi, F., Sassanelli, C., Terzi, S., Taisch, M.: A systematic literature review on data and
information required for circular manufacturing strategies adoption. Sustainability, 13(4),
2047 (2021).

13
18. Dellermann, D., Ebel, P., Söllner, M., Leimeister, J. M.: Hybrid intelligence. Business &
Information Systems Engineering, 61, 637-643 (2019).
19. Jarrahi, M. H., Askay, D., Eshraghi, A., Smith, P.: Artificial intelligence and knowledge
management: A partnership between human and AI. Business Horizons, 66(1), 87-99
(2023).
20. Plaumann, L.: Building resilience after COVID-19: EU measures to protect jobs and pro-
mote skills. Available online: https://www.eurofound.europa.eu/publications/arti-
cle/2023/building-resilience-after-covid-19-eu-measures-to-protect-jobs-and-promote-
skills. Last accessed: 15/06/2023.
21. Ambrogio, G., Filice, L., Longo, F., Padovano, A.: Workforce and supply chain disruption
as a digital and technological innovation opportunity for resilient manufacturing systems in
the COVID-19 pandemic. Computers & Industrial Engineering, 169, 108158 (2022).
22. Thiebes, S., Lins, S., Sunyaev, A.: Trustworthy artificial intelligence. Electronic Markets,
31, 447-464 (2021).
23. Conversational AI Market. Conversational AI Market by Offering, Conversational Interface,
Business Function (Sales & Marketing, HR, ITSM), Channel, Technology, Vertical (BFSI,
Retail & eCommerce, Healthcare & Life Sciences) and Region – Global Forecast to 2028.
Available online: https://www.marketsandmarkets.com/Market-Reports/conversational-ai-
market-49043506.html?gclid=Cj0KCQjw0PWRBhDKARIsAP-
KHFGh6B5mjeTgniXB0CG6HkceFBDGMDxe0O9HbeqJEWG0scGsKD6UQd1MaAquZ
EALw_wcB. Last accessed: 15/06/2023.
24. Asher, N., De Lara, L., Paul, S., Russell, C.: Counterfactual models for fair and adequate
explanations. Machine Learning and Knowledge Extraction, 4(2), 316-349 (2022).