Four Key Trends of Generative AI

AI-powered quality inspection revolutionizes manufacturing by automating product validation while maintaining rigorous quality standards.Welcome to Spotlight AI! Here’s your update on industrial AI. This month, we cover four key fields of generative AI, from agentic AI to specialized hardware, and how these technologies will impact industries. We also include the latest industrial AI news, an insightful report on artificial intelligence in manufacturing, and the podcast of the month. Enjoy the read!

▶️ Listen to this newsletter · 12:18 min

From Hype to Impact: The four domains driving industrial transformation with GenAI

Over the past decade, AI has been delivering real value to industries: from predictive maintenance to generative design. GenAI is evolving rapidly, and with that, new opportunities for industries are evolving. Which are the innovation fields to watch? This article explores four key domains where GenAI is fueling the next wave of industrial intelligence.

Agentic AI: autonomous systems that think and act for industry

Agentic AI, in combination with GenAI, is opening a new era in autonomous and adaptive intelligence.  AI agents operate independently and make decisions on behalf of human operators. They interpret intentions, enhance their skills via ongoing learning, and utilize external tools and other agents when necessary.

"With our Industrial AI agents, we're moving beyond the question-answer paradigm to create systems that can independently execute complete industrial workflows," says Rainer Brehm, CEO Factory Automation at Siemens Digital Industries. "By automating automation itself, we envision productivity increases of up to 50% for our customers – fundamentally changing what's possible in industrial operations."

Consider a factory machine malfunction: while a generative AI Smart manufacturing in action: AI-powered quality inspection systems automatically validate products on the production line, streamlining operations while ensuring consistent quality standardsSmart manufacturing in action: AI-powered quality inspection systems automatically validate products on the production line, streamlining operations while ensuring consistent quality standardsI-powered quality inspection revolutionizes manufacturing by automating product validation while maintaining rigorous quality standards.I-powered quality inspection revolutionizes manufacturing by automating product validation while maintaining rigorous quality standards.assistant requires a technician to manually input error codes and interpret solutions, an AI agent automatically detects issues, diagnoses problems, and creates repair tickets with detailed instructions. This automation reduces human intervention to just the repair work itself, demonstrating how agentic AI can independently handle complex, multi-step industrial challenges.

An “Agentic Economy” will emerge as a marketplace where AI agents can be purchased and engage in direct interaction with one another. These agents will then operate in networks under a “master agent”, with standardized descriptions enabling inter-agent discovery and collaboration. Within these networks, a key challenge lies in maintaining consistent quality control across the diverse range of agents provided by different vendors

Multimodal LLMs: merging text, visuals, and data for deeper insight

Multimodal large language models (LLMs) combine text, images, and video to help machines better understand and interact with the real world. They can interpret visual content, follow complex instructions, and improve tasks like object recognition and scene analysis. This versatility makes them useful in areas like robotics, autonomous vehicles, and industrial automation.

Just as conventional LLMs improved text and speech processing, multimodal LLMs are set to change how industry-specific data is processed: from time series analysis to 2D/3D modeling and machine vision applications.

Building these models presents inherent challenges, primarily due to the demand for extensive, meticulously labelled datasets and the potential for bias. Consequently, a promising solution lies in comprehensive pre-training, which minimizes the effort required for end-user application, thereby enabling easier, safer, and more scalable adoption.

Edge AI: real-time intelligence where it matters most

Edge AI is a growing trend in industrial settings. It means placing AI systems close to where data is generated, like on factory floors or within machines, rather than sending everything to the cloud. Keeping data and AI algorithms within local networks strengthens security and data protection, avoiding the exposure risks of cloud-based processing.

Edge AI allows faster decision-making: data is processed in real-time at the source, reducing latency.  These advantages make Edge AI particularly valuable in scenarios where speed and efficiency are critical, such as predictive maintenance or real-time quality control. On top of that, the edge setup saves huge costs for continuous cloud connectivity.

Technologies like federated learning enable models to be trained directly on multiple local devices. This approach avoids sharing sensitive raw data, thereby protecting privacy while still enhancing overall model performance.

Specialized hardware: powering AI at the speed of industry

Staying at the edge: specialized hardware, like graphic processing units (GPUs) or language processing units (LPUs)-enabled edge devices, is an emerging trend in industrial AI. These devices process AI workloads directly at the edge, enabling real-time analysis without relying on cloud infrastructure. This way, they boost the performance of AI systems by handling complex tasks faster and more efficiently. This is crucial for applications that need immediate responses, such as robotics or autonomous systems.

One standout example is Groq’s LPUs, which are optimized for running LLMs efficiently. Their platform, launched in 2024, is already used by almost 1.5 million developers. 

To ensure seamless integration and optimal performance, careful consideration must be given to factors including power consumption, scalability, and compatibility with existing systems

What’s the impact on industries?

Experts predict that by 2030, industrial AI is expected to evolve from assistance systems to fully autonomous operations. In manufacturing, AI systems will independently monitor, analyze, and control complex processes in real-time, making split-second decisions to optimize operations without any human intervention. This transformation requires building trust in AI's performance and reliability, as manufacturers must be confident in delegating control to autonomous systems that can handle highly flexible, customized, and high-speed processes.

EU enforces new rules for providers of general-purpose models

The European Commission has published new guidelines outlining obligations for providers of general-purpose AI (GPAI) models under the AI Act that came into effect on August 2. Providers are required to ensure transparency, share model information with downstream developers, and comply with EU copyright law. Providers of high-impact models must also assess and mitigate systemic risks. This is a measure to provide legal certainty to actors across the AI value chain, says a press release from the European Commission. Exemptions are also possible if providers meet strict transparency and documentation obligations. Learn more: European Commission. 2-minute read.

Chinese AI firms establish alliances to boost domestic ecosystem amid US curbs

In reaction to tightening US export restrictions, Chinese AI companies have launched two major alliances to strengthen domestic AI ecosystem and reduce reliance on foreign tech. Announced during the World Artificial Intelligence Conference in Shanghai, the “Model-Chip Ecosystem Innovation Alliance” links LLM developers with local chipmakers like Huawei and Biren, aiming to integrate the full AI tech stack. A second group, the Shanghai General Chamber of Commerce AI Committee, focuses on industrial AI adoption. The measures are part of China’s push for self-reliance in AI infrastructure as US restricts export of certain advanced know-how to Chinese companies. Learn more: Reuters. 2-minute read.

Survey: global enterprises rapidly embracing agentic AI across sectors

Agentic AI is gaining momentum in various industries, with 83% of global enterprises viewing it as critical for a competitive edge. That’s according to American analytics company Cloudera’s recent survey of nearly 1,500 IT leaders across 14 countries. The findings show agentic AI is already being deployed in manufacturing for supply chain optimization, quality control, and predictive maintenance; in healthcare for diagnostics, scheduling, and medical records processing; and in retail for customer support, price optimization, and demand forecasting. The survey indicates that an overwhelming 96% of respondents plan to expand their use of AI agents within the next year. Half of them aim for a organization-wide adoption. Learn more: Cloudera. 5-minute read.

The Industrial AI Podcast's July 2025 episode "How Audi uses LSTM in resistance spot welding" showcases a real-world AI application. Hosts Peter Seeberg and Robert Weber, along with guests from Audi and Fraunhofer Institute, explain how LSTM neural networks monitor more than 5 million daily welding spots, transforming quality assurance from static to dynamic. The episode offers valuable insights for both technical and business audiences on implementing AI in manufacturing environments. Click here to listen.

KPMG's report "Intelligent Manufacturing" offers a window into how manufacturing leaders across eight countries perceive and implement AI. Based on insights from 183 senior AI executives, you'll discover how your peers are approaching this technology: 93% see AI as a competitive necessity, 96% report operational improvements, and 62% are achieving ROI above 10%. The report reveals common challenges your counterparts face: 56% struggle with data management, while 78% prioritize sustainability over AI initiatives. You'll also learn how other manufacturers are developing their workforce, with 80% investing in AI skills training. This research provides a valuable benchmark to compare your own AI strategy against industry trends and see where you stand in the manufacturing AI landscape. Download the report

Thank you for reading! In this edition, we explored the four technologies shaping GenAI's future in industry. These trends are boosting performance and expanding possibilities. At Siemens, we’re building on these trends. Stay tuned as we push GenAI's limits and share your thoughts in the comments.

Siemens Editorial Team