The Tactile Expertise of Veteran Workers Is Becoming the AI Foundation of Factories

After a production line completes a product change, it’s time to restart. An experienced operator standing in front of a machine relies on more than just manuals to adjust parameters.

He knows this batch of material will feed more tightly at the current humidity, remembers a minor jam that occurred at 3 a.m. during the previous shift, and can distinguish between normal screen fluctuations and those requiring equipment inspections.

Operational procedures outline step sequences, but a successful first startup depends on experiential judgment and decades of factory-accumulated know-how.

All these elements together constitute industrial intelligence: documented processes, frontline workers’ tacit knowledge, historical production data, and a company’s understanding of its operations.

Most factories face a problem: this intelligence is scattered. What’s on paper stays in file cabinets, what’s in workers’ heads leaves with them, and system-stored data can’t communicate with each other.

The opportunity is clear now: enable decision-making roles to access this intelligence, then connect it back to physical production. The strategic direction is never about building more intelligent agents; turning existing business intelligence into a profit growth engine is the true goal.

How Individual Experience Becomes a Factory-Wide Advantage

When knowledge is trapped in one person, one system, or one factory, its reach is limited. Managers must expand the scope of experience without losing the context that makes the knowledge effective.

When an engineer evaluates a process change, pulling up production records is not enough—they need to know the circumstances behind those records: equipment configuration, quality standards, and tradeoffs operators made between speed and yield. Connecting these pieces lets remote factory teams reuse the experience.

When individual experience is widely accessible, innovation flows naturally to maintenance shops, planning rooms, engineering teams, and production lines. More people can leverage the organization’s knowledge and add new insights.

Solving knowledge access alone isn’t enough. Manufacturing is a physical industry, so all decisions must tie back to materials, machines, and products.

When Intelligence Can Act Directly

Physical AI refers to systems that perceive environments, reason, and execute physical-world actions. Connecting equipment and processes lets factories clearly see shop floor events. Intelligent agents analyze above the data layer, identifying patterns, predicting faults, and calculating actions. Physical AI brings business intelligence into production processes, letting machines and systems adjust automatically to conditions.

A robotic feeding station encounters a misaligned part. The vision system determines the part’s actual orientation, calculates a suitable path using material properties and task models, and the networked control system adjusts within a pre-verified safe range. Value isn’t just spotting deviations; the system must provide appropriate responses.

The same applies to quality inspection. Toyota Industries broke down data silos in its painting shop, letting engineering and operations teams access full context for defect root causes. Public company data shows analysis cycles shrank from 5 days to under 4 hours, and related paint defects fell ~25% in pilots. The breakthrough wasn’t just identifying an extra signal—it was understanding why defects occurred, letting teams act before outcomes solidified.

This is industrial intelligence’s irreplaceable role: data explains what’s happening, industrial intelligence clarifies what should happen, which constraints can’t be broken, and when uncertainty needs human input.

Simulation provides a buffer. Plans run first in digital environments to test impact on processes—this doesn’t replace physical validation, but eliminates using real production lines as test beds.

Responsible implementation starts with operational discipline. Managers must define system authorities, equipment reliability protocols, safety boundaries, and decision/action tracking. Humans must recognize uncertainty, intervene anytime, and bring equipment to safe states. Trust comes from these concrete arrangements, not just model confidence.

As Machines Become More Capable, Humans Must Too

Manufacturing has always blended human creativity and powerful tools. Physical AI extends this: intelligent devices handle suitable repetitive and heavy tasks, while agent-assisted operations help humans process information-heavy decisions and variable conditions.

The goal isn’t just smarter machines—only when knowledge flows and intelligence acts do human ideas have broader room to land.

Rockwell Automation piloted this at a Singapore factory. Technicians used a maintenance assistant tool integrating production data, machine manuals, and senior workers’ experiential knowledge. New hires’ troubleshooting time for hundreds of machines dropped from 9 months to 3. Engineers could validate multiple process options before finalizing changes, and operators stopped repeating the same anomalies while optimizing system responses.

This progress needs investment in both equipment and people: training, learning time, and system design involvement. As machines advance, humans must too.

Industrial Intelligence as a Driver of Long-Term Growth

Industrial intelligence connects a manufacturer’s assets: employee experience, operational signals, and accumulated lessons. Physical AI gives these intelligences a channel to act physically, turning knowledge into decisions into actions.

The industry’s next challenge isn’t just more output—it’s growth without proportional complexity. Intelligence turns experience into scalable capabilities: boosting productivity, accelerating innovation, expanding capacity, and solving once-intractable problems.

The next industrial frontier isn’t smarter machines or fully automated factories—it’s companies that continuously learn, adapt, and act. The opportunity is turning intelligence embedded in every operation into compound-growth advantages.

Watch the series to learn how industrial customers unlock intelligence across value chains, or follow industrial intelligence updates at Microsoft Ignite.

This article was produced by Microsoft and not edited by the MIT Technology Review editorial team.