The AI Hourglass: How Artificial Intelligence is Reshaping Industrial Automation Profit Pools

Session: From Pyramid to Hourglass: How AI is Reshaping Industrial Automation Profit Pools

Speaker:

• Adrien Bron: Partner and Leader of Advanced Manufacturing & Services for DACH, Bain & Company

• Michael Schertler: Senior Partner, Bain & Company

• Timo Kistner: EMEA Industry Lead for Manufacturing and Industrial, NVIDIA

• Dr. John Markus Lervik: Founder, Cognite

• Rainer Brehm: COO for Automation and CTO, Siemens Digital Industries

• Florian Dörrfuß: CTO, Schaeffler Special Machinery

Session Summary: This panel explores the structural transition of industrial automation from a hardware-heavy "pyramid" to an AI-driven "hourglass," where software, data platforms, and smart edge devices hold the highest value. The experts emphasize that achieving closed-loop, autonomous factory operations relies heavily on implementing standardized data knowledge graphs, accelerated computing, and rigorously simulated physical AI. To successfully scale these operational gains, manufacturers must deploy interoperable technologies within existing "brownfield" environments and foster open digital ecosystems.

The automotive and manufacturing sectors are undergoing a profound structural transformation driven by climate targets, disruptive technologies, and shifting global value chains. For decades, value creation in industrial automation was concentrated in traditional control systems, but the integration of artificial intelligence (AI), advanced software, and data platforms is fundamentally changing this logic. Profit pools are shifting rapidly, and competitive positions across the industrial stack are being redefined as intelligence moves deeper into design, production, and operations.

Industry leaders recently gathered to dissect these structural changes, highlighting the transition from rigid legacy systems to dynamic, AI-driven manufacturing environments.

The Structural Shift: From Pyramid to Hourglass

Adrien Bron, Partner and Leader of Advanced Manufacturing & Services for DACH at Bain & Company, explained that the market is entering a new era characterized by closed-loop automation and continuous feedback from operations to design. This shift is fundamentally altering where value is captured.

• The Demise of the Automation Pyramid: Historically, industrial value resembled a pyramid, with the bulk of value residing in field-layer machines and middle-layer control systems.
  

• The Rise of the Hourglass: By 2030, this structure will look more like an hourglass. Immense value will be created at the top (enterprise software, data platforms, and AI workflows) and at the bottom (smart IoT endpoints and edge devices).
  

• The Squeeze on the Middle: Traditional Programmable Logic Controllers (PLCs) and Distributed Control Systems (DCS) in the middle layer are facing immense pressure as their functionalities migrate upward to the cloud or downward to the smart edge.
  

• Real-World Closed-Loop Automation: Bron illustrated this with a connected car scenario. If a vehicle detects a minor braking anomaly on the highway, telemetry data is sent to the manufacturer, who can instantly trace it back to the exact factory and assembly cell. The manufacturer can then proactively adjust assembly tolerances and instruct suppliers to redesign the brake pad before the anomaly becomes a driver malfunction.

The Data Bottleneck and the Need for Context

While AI promises unprecedented operational autonomy, it cannot scale without a highly structured data environment. Dr. John Markus Lervik, Founder of Cognite, emphasized that the primary bottleneck to unlocking industrial value is not the AI algorithm itself, but the underlying data foundation.

• Data as a Prerequisite for ROI: Even prior to the generative AI boom, a properly managed industrial data foundation increased the return on investment (ROI) for digital programs by 300% to 400%. With generative AI, scaled adoption across factories is impossible without trustworthy data.
  

• Knowledge Graphs Over Federated Architectures: The industry increasingly recognizes that data must be organized in contextual "knowledge graphs" rather than standard federated architectures, which fail to provide the high performance and low latency required for autonomous operations.

Accelerated Computing and Physical AI

To realize the potential of "goal-based" automation—where operators simply give a system a physical objective rather than writing lines of code—immense computing power is required. Timo Kistner, EMEA Industry Lead for Manufacturing and Industrial at NVIDIA, detailed how accelerated computing enables this shift across the full AI stack.

• Accelerating the Value Chain: Accelerated compute encompasses physical GPUs, software frameworks, and libraries. For example, fluid dynamic simulation tools, such as Siemens' Simcenter STAR-CCM+, that previously took weeks to process can now be completed in hours using NVIDIA GPUs.
  

• The Three-Computer Model: Kistner outlined NVIDIA's deployment strategy for physical AI, which relies on three distinct computing environments: an AI training ground, a highly accurate physics-based simulation environment to rigorously test models before physical deployment, and an edge computing deployment layer.
  

• World Foundation Models: Because testing unverified AI on physical robots is dangerous and real-world failure data is scarce, Kistner advocates for "World Foundation models". These models intrinsically understand physical laws and can safely synthesize millions of realistic training scenarios.

Deploying in the Real World: Open Ecosystems and Brownfield Realities

Despite these technological leaps, manufacturers cannot afford to abandon their existing infrastructure. Rainer Brehm, COO for Automation and CTO at Siemens Digital Industries, stressed the importance of open ecosystems and backward compatibility.

• The End of Proprietary Systems: The era of closed, proprietary field buses is over. Platforms like Siemens Xcelerator utilize open IT standards, such as Docker-based containers, allowing partners to securely deploy applications directly onto the deterministic shop floor.
  

• Brownfield Integration: New AI systems must seamlessly connect to legacy "brownfield" environments, integrating with standard PLCs via protocols like Ethernet/IP and MQTT to bridge shop-floor data to cloud platforms like AWS and Microsoft.
  

• Standardization Initiatives: Brehm noted that foundational data standards already exist, pointing to the "Asset Administration Shell" developed during the German Industry 4.0 movement over a decade ago; the current imperative is widespread execution.

Florian Dörrfuß, CTO of Schaeffler Special Machinery, reinforced the uncompromising operational realities faced by manufacturers on the factory floor.

• ROI and Zero Downtime: Corporate leadership demands hard efficiency metrics, and AI implementations cannot cause operational downtime because running machines actively generate revenue. Technologies must be highly reliable and readily accepted by floor workers to succeed.
  

• The Burden of Heterogeneous Data: Dörrfuß identified heterogeneous data quality as the single largest barrier to scaling AI. His ultimate vision is an industry-aligned, standardized data foundation that enables end-to-end digital process chains. This would allow real-time production data to feed into a digital twin, enabling engineers to simulate and optimize machine parameters without halting physical production.

The Path Forward

Industrial automation has entered a phase where factories are evolving into highly adaptive systems capable of sensing, learning, and acting autonomously across the entire value chain. As Michael Schertler, Senior Partner at Bain & Company, concluded, these productivity gains are very real and highly scalable. However, realizing this potential requires orchestrated ecosystems, as no single company can build the entire stack alone. The organizations that move quickly to adopt unified data models and accelerated compute infrastructures will dictate the rules of this new era and capture the industry's shifting profit pools.