Who owns the failure when an AI system acts without a human in the loop? How do manufacturers overcome workforce resistance that 25.7% of CPG companies cite as a barrier to AI scaling? Neil Smith of Schneider Electric and Andrew Waycott of TwinThread answer six audience questions on AI ROI, shopfloor training, knowledge transfer, accountability, cultural resistance, and workforce augmentation asked during the “From AI Pilots to Profit: Unlocking Manufacturing’s Hidden Data Advantage” session at the Industrial AI Summit 2026.
Why Does Human-Centered AI Produce Higher ROI?
Audience question:”Human-centered AI is very important. What are your thoughts on ROI + The Person? The actuation loop with person-centered capabilities and needs is important.”
Andrew Waycott, President and Co-founder of TwinThread:
Industrial AI is only as powerful as the experts using it. When we evaluate “ROI + The Person” in TwinThread, we are ultimately measuring how effectively we scale the knowledge of your top engineers and operators. The highest ROI comes from augmenting Subject Matter Experts. Giving engineers an accessible, complete Industrial AI platform eliminates manual data tasks and frees them for proactive, strategic work. Every time they interact with the platform, the system captures that tribal knowledge and self-optimizes. The person becomes the actual driver of the ROI.
This philosophy anchors the actuation loop. When the system detects process drift, it doesn’t blindly execute a change with full autonomy. Instead, it delivers a targeted, actionable recommendation to the operator on the floor. The operator then approves, modifies, or rejects the action, and the model learns from that choice. Ultimately, it is a continuous, self-reinforcing feedback loop that bridges machine intelligence with human domain expertise.
Neil Smith, Segment President, Consumer Packaged Goods at Schneider Electric:
I would start by saying that organizations seeing the best AI ROI are often the same organizations seeing improvements in workforce engagement, retention, and capability, so a wholistic people centric approach is working. Our “Beyond the Hype: Practical AI for Competitive Consumer Goods Manufacturing” paper sets out our vision on this. Industrial AI is not about automation alone, but about meaningful augmentation that uses technology to strengthen human capability, support safety, and improve job satisfaction. Industrial AI should act as decision support, keeping strategic oversight and all ethical considerations fully in human hands. We see agentic AI systems supporting human workers by aligning actions with real-time environmental data and organizational objectives, and by helping them handle complexity and routine decision-making, so people can focus on strategic and creative work. On the ROI side, the same logic applies: employees must be trained to leverage new industrial AI solutions. We must treat workforce capability as part of the ROI equation, not separate from it.
Will Shopfloor Workers Train AI Models in the Future?
If people drive the ROI, the next question is whether shopfloor workers will eventually shape the models themselves.
Audience question:”Do you think that shopfloor workers will train models in the future?”
Andrew Waycott, President and Co-founder of TwinThread:
They already do. While frontline operators do not build or train TwinThread models directly, their historical actions define the initial setpoints and operating boundaries the AI relies on. Operators then drive the continuous AI feedback loop. Every time an operator accepts, modifies, or rejects a recommendation, they generate vital data. Engineers then act as a conduit, using this real-world feedback to iterate on the model, refine it, and push updated recommendations back to the floor.
Can AI Bridge Veterans and Digital Natives?
Operators already shape AI through their feedback. A related challenge is transferring decades of experience from veteran experts to a younger, digitally native workforce.
Audience question:”Combining AI with experienced experts and young digital natives to bridge knowledge. What do you say?”
Andrew Waycott, President and Co-founder of TwinThread:
This is exactly what we encourage at TwinThread. Tribal knowledge shouldn’t be held by a handful of veteran experts. By embedding that operational expertise directly into AI models and digital workflows, we make it instantly accessible to a newer, tech-fluent workforce. This ensures that regardless of turnover or shift changes, the plant’s most critical knowledge remains secure and universally available.
Neil Smith, Segment President, Consumer Packaged Goods at Schneider Electric:
Industrial AI solutions will undoubtedly help capture existing enterprise expertise and pair it with newer talent. As agentic AI becomes more capable at handling complexity and supporting routine decision-making, experienced workers can spend more time focusing on judgment calls, while AI also lowers the entry barrier for digitally native engineers. AI has the potential to become the mechanism through which decades of operational experience are transferred to the next generation.
Who Owns the Failure: Data Engineer, Model Provider, or Operator?
Audience question: “When an AI system takes an action with no human in the loop, who owns the failure: the data engineer who built the pipeline, the model provider, or the operator who trusted the system?”
Andrew Waycott, President and Co-founder of TwinThread:
I don’t believe the fault lies with any single part of the process. It’s a failure of the workflow as it was constructed. A robust system requires shared safeguards across every level. The data engineer building the pipeline should have measures in place to guarantee a clean flow of reliable data. The model provider must establish thresholds to automatically alert the team when a failure is imminent. Finally, engineers and operators need clear procedures to spot anomalies, diagnose issues quickly with the tools at hand, and execute corrective actions before an asset or process is disrupted.
How Do You Overcome Operator and Management Resistance to AI?
Shared accountability requires trust, but building that trust is difficult when operators and middle management feel threatened by the technology.
Audience question:”How do you recommend manufacturers handle the organizational and cultural resistance from operators and middle management who may feel threatened by AI systems? What specific change management strategies have you seen actually work in practice?”
Neil Smith, Segment President, Consumer Packaged Goods at Schneider Electric:
Workforce resistance is a top barrier to AI scaling identified by our 2026 AI in CPG study (25.7% of CPG manufacturers globally cite it as a risk). It’s important to remember that industrial AI isn’t just an IT project. It succeeds when teams work together, leaders provide clear support, and people understand that AI is there to assist them, not replace them. Change management is crucial to bringing people along the journey. A practical suggestion is to engage the operations and maintenance teams early in the process when identifying the opportunities for AI use cases and keep them engaged through the deployment process. The programs that succeed pair platforms and tools with equal investment in change management and workforce capability. Building that capability alongside the technology is what turns a compliance obligation into a lasting operational advantage.
Where Is the Balance Between AI and Employee Redundancy?
The resistance question leads to a broader concern: if AI takes over more tasks, is keeping additional staff simply a way to manage the risk?
Audience question:”Employee redundancy seems an AI risk mitigation. Where is the balance?”
Neil Smith, Segment President, Consumer Packaged Goods at Schneider Electric:
Our “Beyond the Hype: Practical AI for Competitive Consumer Goods Manufacturing” paper sets out our approach to this: industrial AI is not about automation alone, but about meaningful augmentation that uses technology to strengthen human capability, support safety, and improve job satisfaction. The goal should not be fewer people. The goal should be more productive, more capable people, the power of this approach is showcased in our Wuhan Lighthouse factory: agentic AI identified and tracked skill gaps and assigned personalized training, supported by dedicated career paths. Workforce readiness jumped from 20% to 76%, and 56% of employees were upskilled. AI was used to automate repetitive tasks, freeing engineers for high-value work and shortening new-product-introduction cycles by 66.7%, cutting lead time from 36 months to 12. Smart guidance and mentor-pairing reduced technician turnover from 48% to 6%. The same number of workers are employed by the factory, while the output increased and turnover markedly decreased.
These questions were submitted by attendees of the “From AI Pilots to Profit: Unlocking Manufacturing’s Hidden Data Advantage” session at the Industrial AI Summit 2026. Answers provided by Neil Smith of Schneider Electric and Andrew Waycott of TwinThread. The session also featured François Coulloudon of TEEPTRAK Groupe, and it was moderated by Kevin O’Donovan. The session was sponsored by TEEPTRAK and TwinThread.
