Skip to Content

GenAI for Intelligent Industry: a new revolution for R&D and operations

Charlotte Pierron-Perlès, Alex Marandon, Hugo Cascarigny, Yasmine Oukrid
Jun 14, 2024
capgemini-invent

Generative AI’s (GenAI) power to transform every aspect of our lives is now common knowledge

But business leaders are still unsure exactly how to make this revolutionary technology an integral part of their activities. The truth is that the integration of Generative AI for operations requires a very different approach than the integration of traditional AI. This is evident in the sphere of R&D and operations.

In recent years, AI has demonstrated concrete impact across the entire operations value chain. However, the most common and widespread use cases consistently focus on optimizing core processes. One notable development is the way industry leaders use AI to optimize the “physical” manufacturing and delivery of goods. Significant changes include time series analysis to improve process yield or scrap rate, operational research optimizing goods inventory norms, flows of goods or transportation, and computer vision to detect non-conformities.

With its unmatched ability to navigate, digest, and interact with unstructured information and documentation, GenAI reoriented the focus from the “physical” to the “information” world. This means a shift from sensors and connected assets data to documents, which in turn leads to a retreat from manufacturing and supply chain core processes to R&D and industry enabling functions (e.g., sourcing, maintenance, quality and regulatory, etc.).

With this new paradigm in mind, leaders in operations see a new world of opportunities opening up before them, all powered by GenAI .

Key capabilities for R&D and operations

Generative AI for research and operations is a gamechanger for all industries worldwide, optimizing organizations by automating data analysis and customer service tasks. Right now, many organizations are rightly experimenting, developing best practices, and identifying scalable solutions. We at Capgemini Invent were some of the first movers in this space, using our expertise to envision applications.

GenAI for R&D and Operations infographic
Six GenAI capabilities that will make a difference in operations

Even though we are only at the dawn of this transformation, several concrete use cases are already shining bright with more emerging every week. All that remains is to explore the untold opportunities.

GenAI for R&D and Operations infograhic
GenAI for operations: use cases already identified embrace the whole operations value chain

Smart products

The ability of GenAI to emulate a human conversation makes it a prime candidate for enriching user experience with “companion apps”. These apps leverage data collected by connected devices and expose it through a virtual assistant, enabling interactions in natural language and access to insights generated in real time. In the near future, we foresee the emergence of autonomous, edged Large Language Models (LLMs) embedded directly within products, enabling a new range of usage.

Engineering and R&D

As it is probably the most document-intensive area of the industry, GenAI offers a multitude of opportunities for engineering and R&D. The ability of GenAI to digest and synthetize complex information, combined with Retrieval-Augmented Generation (RAG) , enables engineers to easily search and query knowledge base or technical documentation using natural language. LLMs have multiple applications all along the technical documentation lifecycle, accelerating creation of a draft, proofreading or consistency checks against existing standards.

From the innovative use of AI algorithms for Automated Molecule Design in Drug Discovery to the Fragrance Formulation Generator streamlining perfume creation, the applications are far-reaching and go way beyond LLMs. It’s all down to GenAI’s ability to craft detailed 3D simulation scenarios.

Manufacturing quality and maintenance

Leveraging its abilities to extract, synthesize, and classify information, GenAI can optimize and turbocharge several manufacturing, quality, and maintenance processes. For instance, it can accelerate classification, summarization, search and analysis of quality incidents. It can also automate generation of quality documentation in domains with heavy compliance requirements (e.g., life sciences) and automate the documentation of non-conformities.

GenAI-powered software enables human operators in maintenance or engineering to rapidly navigate through documentation of assets in a targeted way. Furthermore, Generative AI in manufacturing and operations will be instrumental in the consolidation of information from other external sources, such as weather forecasts, market insights, and assessment of geopolitical risks. This capacity to integrate data from multiple sources is why GenAI maximizes maintenance planning, accelerates operations, and improves efficiency.

This is all the truer in the case of distributed operations, where data coming from different systems is often heterogeneous and sometimes inconsistent. GenAI’s ability to automatically harmonize and retreat data from distributed environments to ensure it meets quality standards will be key to enable seamless utilization of these data flows and to de-risk associated field operations.

Supply chain and purchasing

GenAI can also significantly increase productivity of interactions across a network of partners, assets, and inventory. More specifically, GenAI can boost the resiliency and efficiency of supply chains in the following ways:

  • Enriched demand planning: Complementary to “standards” ML-based forecasting, by analyzing various exogeneous data sources (customer reviews, social-media trends, articles, etc.) to better understand demand drivers.
  • Efficient sourcing: For instance, through improved upstream suppliers’ intelligence, by crossing external information regarding the upstream supply chain with the analysis of internal suppliers’ documentation and deriving insights from these data. GenAI can also improve procurement process efficiency by automating transactional tasks involving interactions with suppliers or processing of external documentation (contract analysis, CSR compliance check, automated review of tenders, etc.).
  • Improved Customer interactions: Automating and augmenting customer service and back-office operations, by accelerating search, summarization, classification and processing of trade, logistics and customer claims.

Three detailed use cases

The following three case studies provide an idea of how AI will change R&D and operations in the near future.

1. Search, synthesis, and reconciliation of engineering documentation

In the engineering industry, accelerating development processes while securing the quality on most complex products or major industrial projects are key challenges.

The main difficulties usually encountered are:

i) Limited access to engineering knowledge, and/or inefficiencies inherent to collaboration across various business entities involved in one given project.

ii) Time-consuming – or even occasionally unattainable – compliance requirements (e.g., traceability demonstration standards) generating significant complexity to retrieve and match key information from contracts, engineering specs, test procedures and results.

Generative AI in research is proving to be profoundly transformative. It can be used to research and analyze extensive amounts of documentation, both internal and external, such as engineering standards, design practices checklists, papers, and industry benchmarks, encompassing various formats, languages, and structures. It can identify and extract the relevant information and find the best way to display the requested information. It can also document user feedback, measuring engineers’ satisfaction and contributing to enhanced model performance.

GenAI can also support the reconciliation of information across the V-cycle, by extracting key elements, such as specifications and technical requirements, from technical documents (even when these are not properly referenced). Once retrieved, GenAI correlates the data to improve requirements traceability (e.g., high-level requirements with low-level specifications).

Finally, GenAI can accelerate the generation of test procedures and reports.

2. Smart search for technical documents

Engineering teams face the rising complexity of new or changing regulations, extended commercial and industrial ecosystems, technological constraints, customer expectations, and even their own organizational structure and business process management systems. As such, relevant information is often managed by different stakeholders, widely dispersed, and non-homogeneous in various regards (e.g., format, granularity, languages, etc.).

Engineering teams struggle to manage this complexity. Despite being a critical task, information management is highly time-consuming and potentially a source of new risks or missed opportunities.

GenAI can help human operators face these challenges, by leveraging its ability to research and analyze the extensive amounts of documentation required for maintenance and engineering, both internal and external, such as maintenance report, operating protocols.

Additionally, it can facilitate interactions between human operators and databases by writing and executing queries in natural language, which can prove decisive in accessing certain information.

3. Customer service efficiency

Within the consumer products industry, customer service faces several recurring challenges related to searching, summarizing, and classifying clients claims, as well as process logistics:

i) Customer service processes involve interacting with multiple IT systems and requires communication with various stakeholders to respond to customers’ requests, making it complex and tedious to gather relevant information.

ii) These processes are still largely manual, which is time-consuming and increases the risk of errors. As a result, end-to-end claims resolution can take months to complete, impacting both customer satisfaction and cash flow optimization.

To address these challenges, GenAI can be used at several levels:

  • Data retrieval and synthesis: To search for and retrieve relevant information related to the claim from various data sources, such as invoices and delivery receipts and contracts, all of which exist in different formats.
  • Proposition of insights and validation recommendations: Comparing collected information and received claims, GenAI can swiftly detect inconsistencies, highlight discrepancies, provide insights, assess whether the customer’s claim is well-founded, and make recommendations on potential outcomes to suggest to customers.
  • Document processed claims: Capitalization is of the essence. Documenting processed claims, connecting the claim, its outcome, and the evidence used paves the way for easier information retrieval and decision making, should similar cases arise.

From patterns to trends: Key GenAI considerations

We expect the use of GenAI in all industries to scale up at an accelerated pace in the coming months. Below are some of the more exciting zones of development:

Multimodality refers to the capability of GenAI models to process and generate outputs across multiple types of data, such as text, images, and audio. It facilitates more comprehensive and integrated interactions with human users or with other software, enabling the AI to understand and respond seamlessly to complex inputs combining different modalities.

With the release of ChatGPT4o, multimodality is set to augment and empower human operators not specifically trained to interact with GenAI, such as many blue-collar workers. It will radically improve training, upskilling, safety, and eventually the optimization of processes.
More importantly, multimodality paves the way for intelligent systems to interact autonomously with the physical world. There is currently a major push in R&D to develop a new generation of robots able to communicate with humans via speech and imitated gestures.
In short, multimodality is key to unlocking the full potential of GenAI on the shopfloor, having a major impact on the deployment of Industry 4.0 use cases as it matures.

As of today, GenAI for R&D and operations is still essentially addressed as standalone custom use cases, handled separately from the “legacy” systems that support R&D, Manufacturing, and operations – namely, from Enterprise Resource Planning (ERP), Manufacturing Execution Systems (MES), Advanced Planning and Scheduling (APS), and Product Lifecycle Management (PLM) systems, which are key to efficiently manage the shopfloor and connect it to the rest of the business. But the first initiatives to bridge the two worlds are emerging.

Early initiatives, for example, include the generation of content or of code interpretable by MES systems or Programmable Logic Controllers (PLCs). Another example is the generation of component designs, sub-assemblies, or complex systems, all based on human guidance communicated through prompts. However, these initiatives still lack full and seamless integration within the core systems.
In the upcoming months, we can expect IT vendors in operations (ERP, PLM, MES, APS, etc.) to progressively integrate GenAI embedded features in their solutions, like Microsoft has done with Copilot in its Office suite. Considering how widespread these solutions are, this may in return fuel a wider and more systemic adoption of GenAI within Operations – even if standalone custom deployment of GenAI will probably remain a frequent pattern, for client-specific use cases.

With applications in many industries, GenAI can supercharge design generation in many different contexts: molecule generation, product design, chipset conception, or designing parts and components. For instance, in the automotive industry, GenAI models can support the creation of tire design, considering performance requirements and engineering constraints. The consumer products sector is another interesting example, where combinations of GenAI models are used to accelerate the discovery and selection of optimal formulas for new fragrances.

GenAI can completely automate the creation of 2D or 3D designs, concepts, and product architectures. Moreover, it can supercharge computer-aided design or computer-aided engineering models based on requirements and constraints, decide to launch relevant simulations, analyze their results, and adjust simulations if needed. More importantly, GenAI can do all this while also automatically creating design models, including suggest suppliers and logistics schemes, drafting documentation for in-service, end-of-life, and redesign loops. It can also boost the development of eco-design, by providing designers with the latest international regulations and compliances.

RAG: a powerful tool to improve accuracy and limit hallucinations

Retrieval-Augmented Generation (RAG) is an approach combining two basic capabilities of GenAI (information retrieval and content generation), both proven to be very efficient at producing highly accurate responses.

To put it simply, RAG essentially consists of restricting the search field in which a GenAI model will look for relevant information to answer users’ requests to a given, predefined, and limited set of documents. By doing so, this ensures the model will only look within a curated collection of information, the accuracy and quality of which can be guaranteed. This targeted retrieval is then used to generate a more precise and informative response.

RAG is particularly helpful in limiting or eliminating hallucinations, which are instances where the model might generate incorrect or nonsensical information. This is achieved by grounding responses in data coming from reliable and relevant documents. This approach is currently implemented in a vast range of use cases, in various industries where the reliability of responses is particularly important.

Three vectors for success for GenAI in R&D and operations

To lift the potential of Generative AI in Operations, like for any other digital innovation, companies must integrate the technology in their digitalization strategy, build scalable tools and upskill their staff on how to best use these AI tools in their data ecosystem. But GenAI also comes with its specificities. So, how do you avoid the pitfalls and find the steppingstones?

Even more than standard AI, it goes without saying that GenAI is a technology often easy to implement in a Proof of Concept (PoC), but very challenging to scale up: most advanced customer uses are specific with deeply integrated GenAI with legacy IT systems at the core of operations processes. Thus, business leaders looking at the potential of GenAI need to be very pragmatic, adopting a “fail fast” mindset whilst being prepared to iterate. This is the most efficient way to reach achievable targets.

Sandboxes experimentation can be hugely beneficial on a limited operational scope, decoupled from legacy IT. Here, concepts can be tested without first needing to rethink the entire system and perhaps needing to scrap the lot.

We work with our clients to make digital continuity, standardization, documentation of projects, and unification of data a part of enterprise-wide models and infrastructure. And even though GenAI can query and manipulate massive amounts of unstructured data, this should not be an excuse to curtail or stop this support for the quality and structuring of data.

As for traditional AI, large and high-quality datasets are needed to train GenAI. The quality and accuracy of training datasets determine the models’ outputs, and here, robust data foundations remain a necessity. This is particularly true in distributed environments, where data models from different sites are generally heterogeneous, leading to difficulties in replicating experiments and scaling up.
That is why we absolutely recommend keeping data structured in organized semantic models, such as product lifecycle management models. We also believe in maintaining investments in digital continuity transformation. Building clean, structured, reliable, and federated manufacturing and operations data models will be instrumental in the support for deployment of packaged and generic solutions combining AI and GenAI.

GenAI ecosystem is moving fast, with hyperscalers at the forefront. In this race for technology, one cannot be the best in every category, and business leaders of the digital ecosystem must reflect deeply on where their added value lies. Developing technology internally for each use case may not always be the best solution. In some cases, the cost of developing a custom solution, scaling it up, and maintaining it in the long run will simply be too high compared to the additional value it brings. In this scenario, many look to integrate an off-the-shelf solution developed by a third party.

One year ago, “buy” was not always an option, as software editors were not always able to meet demand for GenAI-powered solutions, and there was no other way than “make”. But things are changing fast: currently, editors are increasingly including GenAI-based features in their solutions. This trend will only gain momentum, as most suppliers of existing solutions dedicated to operations are working to enhance their products with GenAI features. In the short-term, we even expect editors to create environments enabling the construction of GenAI-customized solutions.

Final thoughts:

The GenAI landscape is evolving at a phenomenal pace – too great a pace for some. Barely had ChatGPT 3.5 dropped from media headlines when ChatGPT 4o was released and made free to the general public. With this in mind, it is vital that business leaders stay up to date on the technological roadmaps of software providers. This is the only way to know if solutions already exist or should be customized for a specific need. Additionally, be sure to systematically assess and then monitor the value creation of any custom solution developed in-house, all the while asking yourself one simple question: is it worth it?

Authors

Charlotte Pierron-Perlès

Managing Director of Intelligent Industry, Capgemini Invent
Charlotte Pierron-Perlès is the Managing Director of Capgemini Invent Intelligent Industry global practice. She has been an advisor for the last 18 years in several leading industries when it comes to leverage technologies to create business impact, notably in aeronautic, automotive, life sciences and consumer products. She drives C-suite agenda on R&D, operations performance and transformation, connected products and new services, data and Gen(AI). She advises global leading companies to increase top line, further develop competitiveness and comply with sustainable imperatives. Charlotte is used to shape and drive large business partnerships with clients and partners where innovation, technology, data and GenAI are game changers.

Alex Marandon

Vice President & Global Head of generative AI Accelerator, Capgemini Invent
Alex brings over 20 years of experience in the tech and data space, beginning his career as a CTO in startups and later leading data science and engineering in the travel sector. Eight years ago, he joined Capgemini Invent, where he has been at the forefront of driving digital innovation and transformation for his clients. He has a strong track record in designing large-scale data ecosystems, especially within the industrial sector. Currently, as the Global Lead of Capgemini Invent’s generative AI Acceleration Lab, Alex crafts Gen AI go-to-market strategies, develops assets, upskills teams, and assists clients in scaling AI and Gen AI solutions from proof of concept to value generation.

Hugo Cascarigny

Vice President & Global Head of Data & AI for Intelligent Industry, Capgemini Invent
Hugo Cascarigny has been passionate about AI, data, and analytics since he joined Invent 12 years ago. As a long-time member of the industries and operations teams, he is dedicated to transforming AI into practical efficiency levers within Engineering, Supply Chain, and Manufacturing. In his role as Global Data & AI Leader, he spearheads the development of AI and generative AI offerings across Invent.

Yasmine Oukrid

Senior Manager, Intelligent Industry, Capgemini Invent
Yasmine is a key member of the Intelligent Industry Group Accelerator, where she focuses on defining and executing Intelligent Industry strategies to establish a unique market positioning. She is involved in CxO-level business development, strategic deal shaping, and partnership building. Yasmine supports companies in accelerating their Intelligent Industry digital transformation, addressing challenges related to scaling Smart Factory implementations, software-driven transformation, and utilizing Data and generative AI for operations. Her expertise spans across various industries, with a specific focus on Life Sciences, Automotive, Telco, and High-tech sectors.

    Stay informed

    Subscribe to get notified about the latest articles and reports from our experts at Capgemini Invent