Manufacturers must connect workflows before they can scale AI: Webinar review
Manufacturers are making progress with artificial intelligence, but fragmented systems, inconsistent processes and weak links between operational and commercial workflows continue to prevent many initiatives from moving beyond isolated pilots, experts in a recent Manufacturing Today webinar said.
The webinar, How https://info.manufacturing-today.com/webinarsAI is transforming manufacturing operations from engineering to revenue execution, explored what manufacturers must do to convert AI experimentation into repeatable operational value. The discussion moved beyond individual tools and use cases to examine enterprise integration, agentic AI, workflow orchestration, real-time intelligence and the growing need to connect factory performance with revenue, margins and working capital.
The panel comprised Praveen Kombial, global market leader for sales and solutions at Infosys EdgeVerve; Heather Urbanek, AI product leader at Corning; and Mohit Ahuja, strategy and transformation leader at Caterpillar. Manufacturing Technology Brief editor Mark Venables moderated the discussion.
The gap between pilots and scale
An opening audience poll exposed the distance many manufacturers still have to travel. Only eight percent of respondents said they had successfully scaled AI across multiple operational areas, while a far larger group remained at the pilot or experimentation stage.
That result framed a conversation that repeatedly returned to the difference between proving that AI can work and embedding it into the way a manufacturer operates. A narrowly defined pilot can deliver an impressive result, but the challenge changes when the same approach must work across different plants, systems, product lines and teams.
Years of investment in ERP, MES, PLM, quality and supply chain platforms have given manufacturers access to more technology and data than ever before. Yet those systems have often been implemented around individual functions, facilities or immediate operational needs. The handoffs between them remain fragmented, leaving employees to reconcile information manually and processes to break down at precisely the points where data should move from one part of the enterprise to another.
AI does not remove that complexity simply by being added on top. Poorly aligned processes can be accelerated just as easily as good ones, while a use case designed around the conditions in one facility may struggle when it encounters different data, equipment or working practices elsewhere. The webinar examined why manufacturers need to look beyond the performance of an individual model and consider the workflow, ownership and operational context surrounding it.
Rethinking the role of AI
The discussion also challenged some of the assumptions developing around agentic AI. The term is increasingly used across industrial technology, but the panel drew a clear distinction between genuine intelligence and conventional automation presented under a new label.
Rules-based automation remains valuable where a process is predictable, and the required action is known in advance. Agentic AI promises something broader: the ability to interpret signals, draw information from several systems and coordinate a response across a more complex workflow.
That distinction matters in manufacturing, where an apparent exception in one system may be connected to production constraints, supplier performance, inventory availability or a customer commitment elsewhere. The value does not come from automating another isolated task, but from helping the business understand the wider situation and decide what should happen next.
The webinar explored what this could mean in practical environments, from warehouse operations and supply chain decisions to quality, forecasting and customer order management. It also considered the limits of autonomy. Questions of accountability become unavoidable when AI influences decisions involving safety, compliance, product quality or customer outcomes, making governance and human oversight central to deployment rather than issues to address later.
Connecting operations with commercial value
A second audience poll identified difficulty demonstrating measurable business value as the most significant barrier to scaling AI. That finding shifted the conversation from technical capability to the outcomes that ultimately determine whether a program receives wider investment.
Manufacturers face pressure to improve responsiveness while protecting margins, cash flow and working capital. Operational improvements therefore need to be connected to commercial performance. Faster production has limited value if an order cannot be quoted accurately, inventory cannot be located or a customer request becomes trapped between disconnected systems.
The panel discussed how AI-driven workflow orchestration can create a layer across existing applications, bringing information together without forcing manufacturers to wait for lengthy consolidation programs or replace every legacy platform. This is not presented as a shortcut around data quality or process discipline, but as a way to make progress in the real enterprise environments manufacturers operate today.
Several examples illustrated the scale of the opportunity, including faster warehouse execution, improved order conversion and more responsive product traceability. The full webinar examines how those applications were approached, the results they delivered and why employee experience proved as important as the underlying technology.
Turning information into action
Real-time operational intelligence provided the final thread connecting the discussion. Manufacturers have invested heavily in dashboards and visibility, but seeing a problem does not in itself improve performance. Value emerges when information reaches the right person or system early enough to change the outcome.
The panel considered how earlier signals could help manufacturers respond to quality deviations, supplier delays, labor constraints and changing demand before those issues cascade through production and logistics. It also examined why companies frequently stop at visibility and what must change before intelligence can actively guide execution.
The final audience poll pointed to supply chain and cross-functional decision-making as the areas most likely to benefit from AI-driven orchestration. Yet the wider message was that no single function can scale AI alone. Sustainable progress depends on connecting workflows, choosing use cases around genuine business pain and establishing the governance needed to build trust.
The webinar offers a candid assessment of where manufacturing AI stands today, why so many promising initiatives lose momentum and how companies can move forward without waiting for a perfect technology environment. Watch the on-demand discussion to hear the panel’s examples, practical recommendations and perspectives on what successful AI-driven operations could look like.
