Faster launches start with planning, not commissioning

Ask four manufacturing experts how to speed up a production launch and few will begin with the launch itself. At Automotive Manufacturing North America (AMNA), a panel drawn from Lucid, Ford, General Motors and Magna made the case that the clock starts long before the first part leaves the line.

Arun Prakash Mantena, Global Manager for Advanced Manufacturing Engineering in stamping at Lucid, was blunt about it. “A launch does not start where we are in commissioning. It starts with the planning, with teams breaking down the silos between them and planning from the start.”

Standardise first, then accelerate

The panel’s shared answer formed the entire template. Lucid standardises equipment, controls architecture and processes between its US and Saudi Arabian operations, so that each new launch repeats something that already works. Ford takes the same approach to electrical hardware and PLC software, which cuts the engineering needed for each programme.

Walter Laplante, Controls and Industrial Connectivity Supervisor at Ford, added that the carmaker then builds product changeovers inside a virtual environment, which reduces the programming required on the floor.

Murugan Boominathan, Senior Analyst for Manufacturing Solutions at Magna IT, offered a warning to anyone tempted to treat templates as a shortcut. “speed does not mean that we have to eliminate key principles or key steps.” In his view, collaboration and communication matter even more than the template itself.

Greenfield freedom, brownfield friction

Our panelists made one thing clear; the difference between a new plant and a running one comes down to constraint. A greenfield offers room to optimise, whereas a brownfield must keep producing while the change goes in around it.

Laplante described brownfield sites at Ford full of obsolete components, each one a design decision and a software conversion waiting to happen. His team earmarks six months to a year of upfront engineering and laboratory testing for such sites, while a greenfield is a clean slate that shortens the timeline.

Alberto Ely Vazquez, Senior Process Engineer at General Motors, put the weight on assessment. “The better you do your assessment, the more successful the change will be later on.” This is 'baking in' the outcome from the outset.

Designing for the next change appears to be the other half of the problem, however. Vazquez described configuring equipment like a Lego block, leaving room for flexibility and documenting machines properly so that moving an asset is faster.

Mantena pointed to utilities. In stamping, foundations and services are woven into the building, so they must be planned for needs that do not yet exist. Vazquez drew the logical conclusion. “the greenfields of today will be the brownfields of tomorrow.”

Study China, but do not copy it

Moderator, and AMS Chief Content Officer, Christopher Ludwig asked the question that many automakers may try to avoid, but would be at amiss to do so, that is, whether the panel was chasing Chinese development speed.

Laplante has visited Chinese OEMs and found the same equipment bought from the same limited pool of stamping suppliers. The difference, he said, lies in the standard processes people follow every day. Vazquez was equally unmoved by the mystique.

“I do not see that they have any alien technology in China that we do not have here.” He did admire the tight partnership between Chinese OEMs and suppliers, and argued that North American teams should find their own pace without sacrificing safety, quality or reliability.

Where the digital twin stops

Boominathan said virtual definition now happens at the start. A conveyor can be specified with its stations, routings, failure modes and KPIs ahead of time, so that on the plant floor “We just do the validation.”

Mantena set the limit. Virtual commissioning shows that a product can be made, but “it cannot guarantee that we can produce at rate.” Coil and gauge variation, equipment integration and fault recovery all still need physical proof.

Sitting in the audience, Luv Aggarwal, Portfolio Development Executive at Siemens Digital Industries Software pushed against that view. He argued that synthetic data and agentic capability now let customers validate quality virtually, though no model can predict a miswired connection. Laplante said Ford already uses virtual commissioning to validate and simplify code with generative AI, to teach suppliers its templates, and to program robot recoveries offline.

Vazquez noted that GM now writes and simulates control code while machines are still being built, rather than waiting for them to arrive.

Data first, then the people

Underneath all of this, however, sits connectivity. Laplante set out the sequence, which starts with a robust, secure network, moves to standardised messages through Ford’s Common Data Model, and then allows condition-based monitoring that predicts failures and shortens mean time to repair.

Boominathan agreed that scaling AI of any kind requires a common data model and that older hierarchical integration is fading.

In the closing round, safety and quality were treated as untouchable. Vazquez added a third item, which is training. “a lever is only as strong as its pivot point.” For him, the pivot is engineering fundamentals. A carmaker that wants to launch faster might begin by asking how well its people can use the tools it already owns.