Rethinking AI and Innovation: Transitioning from Output-Driven to Process-Centric

The excitement surrounding Generative AI (GenAI) in industrial environments primarily revolves around specific applications and use cases—what I call an output-driven approach. This perspective resonates with users and stakeholders because the benefits and ROI are tangible. There’s a clear tool, defined CAPEX and OPEX, and visible results.

The simplicity of this approach is appealing: need an industrial co-pilot? Done. Looking for a knowledge management tool? Here it is. Best of all, it doesn’t demand a complete overhaul of existing processes. You can either create a new workflow or seamlessly integrate an AI-powered app into an existing one, such as using it as a recommender system for service desk operators.

But what about using GenAI, machine learning, and automation to rethink and enhance entire processes? This approach views the transformation journey itself as the goal, which often requires process redesign—a familiar exercise for many organizations over the past decade. This time, however, the redesign is supercharged by the technological leap GenAI provides, enhancing the synergy between AI and automation.

Take production planning in an engineer-to-order setup, for instance. These workflows involve complex interactions across order management, material and logistics coordination, and cross-departmental collaboration with sales and customers. Traditional software facilitates smooth information flow and integrates real-time data from operational systems, yielding benefits like first-time-right outcomes, transparency, and customer satisfaction. However, such solutions are often heavily customized to align with the company's unique needs, stakeholders, and systems.

The next evolution involves integrating robotic process automation (RPA) to handle repetitive tasks more efficiently and reliably. AI further amplifies this by optimizing processes—not just to an “efficient” state, but to a “super-efficient” one. Every aspect must be considered: the people involved, data inputs and outputs, and the value delivered to end customers.

For instance, AI chatbots and GenAI assistants can provide decision-making support by offering rapid access to insights and analyses. AI-based collaboration tools can enhance communication and coordination among departments, while AI-driven resource allocation can maximize efficiency and minimize costs. Planners can also benefit from AI-powered simulations to test “what-if” scenarios and refine production strategies.

Beyond optimization, the concept of virtualizing entire processes using digital twins or industrial metaverses is becoming a reality. These technologies allow organizations to simulate production scenarios and explore the best strategies without disrupting actual operations.

Emerging technologies like large language model (LLM) agents are also noteworthy. These agents act as integrators between various process components. However, as Dr. Michael May, Head of Technology Field Data Analytics & AI at Siemens Technology, points out, selecting the right use cases is essential due to the nascent stage of these approaches. Tracing errors in complex workflows remains a challenge.

The Bottom Line

Many organizations remain overly focused on isolated use cases, potentially missing the bigger picture: achieving greater efficiency and resilience through AI-enabled, people-centric processes. Challenges like sparse data, data quality, trust issues, and scaling best practices across factories are real but solvable.

Sparse data can be mitigated with synthetic data and reinforcement learning. Pre-trained models make AI accessible to non-experts. Ultimately, success hinges on collaboration among process owners, end users, optimization teams, technology experts, and systems integrators.

The day when AI designs processes, automates tasks, and trains both models and people is on the horizon. While we’re not there yet, the path is clear—and exciting!