Beyond automation: How Generative AI is redefining manufacturing

In the fast-evolving world of technology, staying ahead of the curve is a harsh reality. Over the past two years, Generative artificial intelligence (GenAI) transitioned from a futuristic concept to a tangible transformative force, shaping the manufacturing landscape in ways previously thought impossible. As global manufacturers navigate the complexities of modern production, GenAI offers powerful solutions to enhance efficiency, increase transparency, reduce costs and drive innovation.

In Deloitte’s recent Future of Manufacturing study—which gathered responses from nearly 600 respondents across manufacturers—a staggering 87% reported that they had already initiated a GenAI pilot. While 24% of respondents indicated that they had adopted GenAI use case(s) in at least one of their facilities, 10% have implemented it across their broader networks. Half of the respondents rated GenAI among the top-priority solutions they planned to implement in the next 24 months, consistently ranking higher than digital twins, the omniverse and the metaverse.

The rise of GenAI in manufacturing

As GenAI moves from hype to reality, let’s take a step back and understand what GenAI is.

GenAI is a subset of AI that utilizes advanced models—including large language models (LLMs) and techniques such as natural language processing (NLP)—to generate new content, designs and solutions. These LLMs allow GenAI to interact with users in a more humanlike conversational manner. Use of knowledge graphs can enhance GenAI’s understanding of language, context and meaning. Overall, GenAI excels in generating new multimodal content, extracting and simplifying data, contextualizing information, and providing a conversational interface.

GenAI capabilities are promising for manufacturers

Top 3 GenAI Capabilities that are most promising to manufacturers

1. Data extraction and simplification

Rapid analysis of large volumes of data to identify patterns and key insights, automated content generation and summarization—along with personalized knowledge delivery—enable GenAI to transform knowledge management from static systems into dynamic, intelligent ecosystems. This empowers the workforce with the right knowledge at the right time, enhancing efficiency and decision-making.

• Workers can quickly access digitized formats of standard operating procedures, manuals, logs, batch records and other documents to help improve operations, resolve queries and make faster decisions.
• Tacit and historical knowledge of technicians can be captured to customize trainings based on specific needs and styles. This helps improve the effectiveness of existing training programs and facilitates faster onboarding.

Furthermore, GenAI models can complement traditional AI prediction models to enhance data analysis by providing richer, more comprehensive insights. This can help manufacturers to improve operations, optimize production planning, minimize out of stocks, predict equipment failures, and analyze product defects.

As the technology evolves, increased adoption of retrieval interleaved generation (RIG) and hybrid retrieval augmented generation (HybridRAG) techniques are expected to further increase value to manufacturers—by enabling higher-quality and domain-aware outputs grounded in a more reliable and comprehensive information source. This is especially beneficial for tasks demanding complex reasoning and precise control over information integration.

2. Context-aware conversational assistance

With its ability to understand human language, GenAI can provide context-aware conversational assistance. Smart systems comprehend the meaning of user inputs and adapt responses based on user preferences, creating natural and meaningful conversations. This GenAI capability can significantly aid manufacturers by providing visibility for performing root cause analysis, enabling them to take the most appropriate mitigation actions.

• If a user asks about optimizing a production line, GenAI-powered applications can consider current production schedules, production constraints, resource availability and past optimizations to provide insightful and actionable recommendations for improvement.
• Operators, supervisors and managers can use the conversational interface to track production and inventory in real time, helping solve exception-based situations. GenAI- powered applications can assist in quickly analyzing the root cause of shop floor incidents and suggest potentially preventive and corrective measures.
• The image below illustrates how GenAI-based assistants can be used to quickly provide a step-by-step approach for resolving a particular error code in a conversational manner. The operators can ask clarifying questions that previously required the intervention and assistance of a tenured operator.
  

Conversational Assistance at Shop Floor at click of a button

3. Multimodal proficiency

GenAI models demonstrate increased versatility in handling diverse data formats. Users can provide input in various modalities, including text, images, audio, code, video and 3D models. Correspondingly, these models can also generate outputs across a similar range of modalities.

• Maintenance workers could input queries using various methods, such as text prompts with error code and/or audio messages with details—even uploading images of malfunctioning parts into a GenAI-based maintenance chatbot. This would use data from diverse sources such as sensor readings, maintenance logs, technician reports and visual information to diagnose equipment faults. The chatbot could generate multimodal outputs, such as text-based troubleshooting instructions, visual aids illustrating repair procedures and even audio overlays with step-by-step guidance for the technicians.
• By integrating diverse data such as sensor readings, visual inspections, audio and maintenance logs, GenAI can enhance machinery fault detection. This can help spot potential issues missed by traditional methods.
• GenAI could also be instrumental in developing immersive learning experiences for workers by creating training modules, process documentation and standard operating procedures from raw text, picture and video feeds. Even the learning modules can be a combination of textual guides, video snippets and real-time simulations—utilizing adaptive quizzes and personalized feedback to maximize impact.

GenAI can help manage challenges in the manufacturing industry

As companies adopt GenAI, proactively addressing challenges related to data privacy, security, availability and quality is paramount. Furthermore, the potential for GenAI model "hallucinations" and the ever-changing regulatory landscape require careful and ongoing attention. Deloitte’s Trustworthy AI™ framework is one such comprehensive framework that incorporates these considerations and supports responsible and sustainable GenAI deployment. Other potential solutions for some of these concerns are to:

• Minimize data privacy risks when using public LLMs by enforcing zero-retention policies with service providers and simultaneously implementing stringent access controls within your own systems to restrict data access and modification privileges related to GenAI applications. This dual approach protects sensitive information both externally and internally.
• Use explainability (XAI) techniques, change management and other forms of trust-building approaches.
• Use guardrails to ensure ethical, secure, transparent and reliable use of GenAI models.
• Establish a GenAI quality assurance team and develop a comprehensive observability platform, incorporating tracing and evaluation capabilities for LLM applications. This platform should support both application-specific benchmarks and real-time performance evaluations, facilitating robust reporting and continuous improvement. Application evaluation frameworks should benchmark the results against bias, accuracy, coherence and relevance.
• Use high-quality data, effective prompt engineering, strong grounding, fine-tuned GenAI models and continuous evaluation to minimize hallucination.

Start small but plan big

GenAI isn’t just a technological advancement—it’s a paradigm shift that’s redefining the manufacturing landscape. It holds the potential to drive significant efficiencies, create new opportunities and solve many age-old industry challenges.

While manufacturers have historically been cautious of adopting process automation, the industry is already moving toward agentic AI wherein intelligent, discoverable and trustworthy AI agents could independently accomplish tasks and make decisions. The adoption of GenAI is now table stakes. 

Identification and adoption of high-value use cases will be critical for successful large-scale adoption. By starting with small, strategic implementations, manufacturers can build a solid foundation for broader transformations. The journey may be challenging, but the rewards can be transformative. The path to more resilient, agile and smarter manufacturing awaits with GenAI.

Author:

Tim Gaus Principal Deloitte Consulting LLP tgaus@deloitte.com Thank you to our contributors: Kreeti Mahajan and Debashish Chatterjee.