Skip to content
‹ Back to IndustriesIndustrial IoT

Factory AI that tunes machines works best with limits on what it can change

GridRaster co-founder Dijam Panigrahi expects closed-loop quality control, spatial AI and physical AI (adaptive robotics) to shape manufacturing over the next 12 to 18 months, with each enabling software to affect what happens on the plant floor. Parsec’s 2026 report says 72% of manufacturers use AI in some form, but only 10% have deployed it at scale. The piece argues that moving from pilot to production starts with one measurable line, clean time-stamped machine data, clear limits on what systems may change (with operator override and logging), and accountable ownership before go-live.

This story was produced through MarketScale. See how Industrial IoT teams put it to work with AI Visibility (GEO).

By MarketScale Newsroom · Closed-loop Quality ControlSpatial AiPhysical AiGridraster
Share
Factory AI that tunes machines works best with limits on what it can change

Key takeaways

01

Seventy-two percent of manufacturers use AI somewhere and 10% use it at scale; the article argues that closing that gap requires deciding before go-live what each system is allowed to change on its own and who owns the outcome.

02

Bounded parameter ranges, operator override and a log of every automatic adjustment are the guardrails that let closed-loop quality control keep the traceability regulated plants need, and they can be written into a purchase specification today.

Free workspace

Turn your Industrial IoT expertise into content.

Record interviews, organize footage, and write with AI on a free trial of the MarketScale platform for qualifying companies. No demo required, no credit card.

Try it Free

A controller that lowers its own injection pressure in the middle of a run is a different machine from one that sends an alert to a dashboard. That difference runs through all three industrial AI shifts Dijam Panigrahi expects to shape manufacturing over the next 12 to 18 months, in an outlook Automation World published on Nov. 19. His three are closed-loop quality control, spatial AI that keeps a live 3D model of the plant, and physical AI that lets robot cells handle part variation without being reprogrammed. Panigrahi is co-founder and COO of GridRaster, which sells spatial computing software to manufacturers, so the second trend is his company's own field.

The backdrop is a gap that hasn't closed. Parsec Automation's 2026 State of Manufacturing Industry Report, based on a February survey of 1,200 manufacturing leaders worldwide, found that 72% have adopted AI in some form and 10% have deployed it at scale. Two years earlier, 53% had adopted it. Panigrahi adds labor pressure, citing a Deloitte estimate that as many as 1.9 million US manufacturing jobs could go unfilled through 2033 if workforce gaps persist.

Parsec's breakdown shows where the adopters actually sit. Another 22% are actively implementing, 28% haven't started, and the remainder, about 40%, are piloting or in early use. That makes pilots the single largest group, which is why the question of how a pilot becomes a production system matters more than which model a plant picks.

Each of the three shifts lets AI change something on the floor, whether that's a setting, a route or a grip. For a plant operations leader, that adds a task ahead of choosing a model: deciding what the system is allowed to touch.

Quality checks that adjust the machine

Panigrahi's first shift starts before production. Citing aPriori, he puts more than 70% of a part's cost in place before manufacturing and sourcing teams have much room to respond. Software that analyzes a part's geometry can catch thin walls, undercuts and tolerances set tighter than the application requires, and it can estimate cost drivers before anyone quotes tooling. An expensive tooling revision turns into a cheaper conversation with the designer.

This use case never writes to a machine, which makes it the lower-risk place to begin. It would matter most for shops quoting dozens of new part numbers a month, where Panigrahi says automated review also shortens quoting cycles and frees senior engineers for harder work. He gives no figure for how much faster quoting gets, so a buyer would need to measure that on their own part mix.

On the shop floor, the shift is from catching defects after a process finishes to preventing them while it runs. Some closed-loop deployments report scrap costs falling by up to 25%.

In closed-loop quality control, readings from vision systems, metrology and sensors go back to the machine controller, which can change feed rates, thermal limits or injection pressures in real time. The difference from AI-assisted inspection is the write-back path: the system acts on what it measures instead of reporting it to a person.

The guardrails are where an operator can act now. Panigrahi describes the strongest implementations as having three features: defined limits on what the system may change, operator override, and a log of every adjustment. That log keeps the traceability regulated sectors require. All three can go into a request for quotation as plain requirements, and asking for them is our recommendation, not a published standard.

Read the 25% as a ceiling. The outlook doesn't name the processes, sensors or control variables behind it, and a plant whose dominant defect forms faster than its feedback loop can respond won't see that kind of number. The starting point is common, though. Quality control is already manufacturers' most-cited AI use case at 50%, according to Parsec, so for many plants the closed-loop step means adding the write-back path to work they already do.

Live plant models and adaptable robots

Static digital twins, in Panigrahi's account, are rule-driven, expensive to build and brittle when the floor changes. Spatial AI uses cameras and LiDAR to build a 3D model that keeps updating, so a planner can check a proposed schedule or layout change against the floor as it stands today, not a survey taken last quarter. Since the model refreshes whenever equipment is moved or the layout shifts, simulations track real conditions more closely, which he says helps teams decide faster on line balancing and material flow.

The same perception lets mobile robots work in older buildings with cracked floors, uneven lighting and tight aisles, without the beacons or reflectors many plants assume they need first. He points to an automotive site where more than 480 autonomous mobile robots share the floor with hundreds of manually driven forklifts and no floor markers. Intent prediction lets the robots anticipate where a forklift or worker is going and yield instead of stopping at every obstacle.

Physical AI targets the work that conventional robot cells handle poorly: high-mix, low-volume production, where parts and fixtures change often and where, Panigrahi argues, much of manufacturing's value is created. An adaptive cell copes with part variation through perception and learned behavior, rather than needing a fresh programming project each time production changes over. Its baseline skill comes from foundation models trained on large amounts of manipulation data. Integrators then tune that skill for a particular plant, which cuts the time between installation and productive work.

The evidence is still early. Panigrahi points to a pilot at a US automotive plant where a humanoid robot handled 90,000 components, and he describes such deployments as bounded and closely supervised.

That timing matches where buyers are. In an A3 outlook also published Nov. 19, Aaron Prather, A3's director of market intelligence, wrote that North American robot orders rose modestly in the first half of 2026 even as demand from automotive OEMs weakened. Semiconductors and electronics, life sciences, food and consumer goods, and metals made up the difference. Prather notes that work outside automotive tends to involve smaller volumes and more product variety, which is exactly the job physical AI is meant to do.

Those customers also bring different expectations about implementation costs, Prather adds. He sees AI-enabled perception as a possible way to automate processes with more variability than traditional automation can handle economically, which puts the adaptive-cell pitch and the non-automotive order growth on the same side of the ledger.

Prather also gives the strongest case for caution. A system that works in a controlled demonstration can still need substantial engineering, validation and maintenance before it runs reliably in production, and he expects 2027 buyers to favor applications with measurable gains in productivity, quality or flexibility. Manufacturers, he writes, need systems that deliver consistent results, meet safety requirements, integrate with existing equipment and run reliably over extended periods.

For an operations leader, a 90,000-part pilot shows the method can work in a building designed for people. A multi-shift production commitment needs more evidence than that.

Where pilots stall

Panigrahi's diagnosis is that the technology is moving faster than the foundations underneath it. He cites a finding that 48% of industrial respondents name legacy system integration and data silos as a top barrier. Parsec's survey asked the question differently and got different numbers: 38% cite difficulty integrating with existing systems, behind high implementation costs at 40% and data privacy and security at 39%. Both point to the same problem, the equipment already on the floor.

What a scaling plan is up against

69%
of manufacturers run a mix of legacy and modern equipment, so most AI will sit on older controls
37%
have a unified, data-driven strategy in place
45%
cite internal skill gaps as a major challenge when integrating new systems
60%
say IT and tech specialists are their hardest roles to fill

Parsec Automation, 2026 State of Manufacturing Industry Report

The 37% figure is moving. Parsec found another 60% in the implementation or planning phase of a data strategy, up from 40% in 2024. The hardware is spreading as well: 54% of manufacturers list AI/ML-enabled decision support among their top tools, and 50% each name IIoT and edge devices and predictive maintenance tools. Those sensors and data streams are what a closed loop reads from, which suggests many plants are nearer the first step than the 10% at-scale figure implies.

The way out in the Automation World piece is narrow on purpose. Start with one line that has measurable scrap or changeover problems. Build clean, time-stamped machine data before adding models. Before going live, set which adjustments each system is allowed to make on its own, and name the person responsible for the result. Expand only after the narrow deployment has run reliably for several months.

Seventy-two percent of manufacturers use AI somewhere and 10% use it at scale; the plants that close that gap tend to decide, before go-live, what each system may change on its own and who owns the outcome.

Two costs often go unbudgeted. Systems that write to machine parameters widen the attack surface and raise the cost of a bad decision, so Panigrahi wants cybersecurity in the plan from the start. Operators who supervise autonomous systems also need different skills from operators who respond to alarms. Parsec's data shows manufacturers are already short of IT specialists, and 49% say quality assurance staff are hard to hire.

Prather's outlook puts names on the people involved: robotics technicians, controls engineers, maintenance personnel and system integration specialists remain essential to deployment. Writing about small and midsized manufacturers, he adds that evaluating processes, preparing facilities, integrating equipment, training employees and setting up maintenance can make up a substantial portion of the total investment. A budget that prices only the software would miss most of that work.

Before signing for any system that adjusts a machine, ask the vendor to show three things on a live line: the parameter bounds it cannot exceed, how an operator overrides it mid-run, and the log entry each automatic change leaves behind.

Parsec found more leaders worried about being too hesitant with AI (60%) than too aggressive (40%). That pressure could push plants to start more pilots. The method above points the other way: fewer pilots, each run long enough to give a plant the evidence it needs to repeat the result. Parsec CEO Suzanne Rudnitzki lands in a similar place, tying technology success to a strong data foundation, an execution layer and a focus on scaling AI across networks instead of only piloting it.

Orlando in January

The next chance to see whether suppliers agree comes at A3's Business Forum, Jan. 18 to 20 at the Signia by Hilton Orlando Bonnet Creek. Assembly Magazine's JJ Moore reports that the members-only event expects more than 700 automation executives. Its program includes a session with leaders from Beckhoff Automation, Teradyne Robotics and Cognex on 2027 opportunities and challenges, along with A3 Executive Vice President Alex Shikany presenting the association's latest automation statistics.

The demand side gets its own airing. Quality Magazine's listing of the agenda adds ITR Economics' Connor Lokar on the economic forces shaping 2027 and a closing discussion with analysts from A3, Interact Analysis and 3DM Consulting. The room is senior: 69% of past attendees hold director-level or higher roles, and 37% are C-suite executives, owners, founders or presidents.

Two numbers from Orlando will test this reading. One is whether non-automotive orders keep growing into the second half of 2026, the high-mix demand that adaptive robotics depends on. The other is whether supplier executives can point to closed-loop and physical AI systems running unsupervised shifts in production, beyond pilots. If they can, the 10% figure in Parsec's next report should look different.

Featured companies

Your experts belong here

Every story in MarketScale Industrial IoT starts with a company putting its controls engineers, plant-floor specialists, and integration partners on the record. Buyers are already reading this topic. The only question is whose experts they find.

Plant and controls buyers research deep before contact, and your engineers get to shape that research.

Book DemoSee how it works15 minutes, straight to a calendar.

About the author

MarketScale Newsroom
MarketScale NewsroomEditorial Team, MarketScale

The MarketScale Newsroom reports on the companies, technologies, and trends shaping 16 B2B industries. It turns primary sources and expert commentary into clear, useful coverage for the people doing the work.

B2B Weekly

The week in Industrial IoT, and sixteen other industries, every Monday.

Ten stories, one-line takes, five minutes. Free.

Industrial IoT: are you visible to AI?

Before they reach out, Industrial IoT buyers ask AI engines which vendors to trust. Explore how your experts, customers, and partners can become useful content for buyers and AI search.

Free Trial

You just read one Industrial IoT expert. Your company is full of them.

This article was produced through MarketScale. The same platform turns your controls engineers, plant-floor specialists, and integration partners into the articles, video, and social content Industrial IoT buyers are searching for. Start a free trial and see it with your own people. For qualifying companies, no credit card, no demo required.

NPS +73 · 1,000+ creators · 38+ countries

What your free trial includes

Hands-on access to the MarketScale platform
Media requests to your crowd, remote recording, AI writing tools
No demo required. No credit card.
For qualifying companies. Company confirmation required.

More Industrial IoT Insights

Atlas Sayaç opens international R&D project call for water metering and industrial IoT

Atlas Sayaç, a Turkish engineering and manufacturing company, opened an international R&D call on November 19, 2026 for water metering and industrial IoT projects. The company proposes linking payments to project success and revenue sharing, inviting water utilities, equipment manufacturers, systems integrators and researchers to collaborate on engineering problems or technologies ready for commercialization.

  • 01Atlas Sayaç is proposing a success-based R&D arrangement for mutually accepted projects, deferring payment until the project meets agreed success criteria and the partner begins earning revenue.
  • 02The company combines electronics, embedded software, mechanical design, manufacturing and software-platform integration across M-Bus, LoRaWAN, NB-IoT/GSM and RS-485 technologies.
  • 03Prospective partners submit a nonconfidential outline of their engineering problem or technology, application, target market and commercial goals to https://atlassayac.com/en/rd-partnerships/ for feasibility discussion.

Nov 19, 2026

Rockwell gives AI agents only the simulation tasks a validated digital twin can support

Rockwell gives AI agents only the simulation tasks a validated digital twin can support

At Automation Fair 2026, Rockwell Automation laid out rules: AI agents may chain digital twin simulation tools only for tasks a validated model supports. Model validation comes first. Green, yellow and red zones set where virtual tests replace physical ones, starting with one or two engineering questions.

  • 01An AI agent's scope in Rockwell's method is set by how well the digital twin has been validated: green-zone tasks can replace physical tests, yellow-zone tasks can trim some, red-zone tasks are off limits.
  • 02People who use simulation results outnumber the engineers who run simulations by about 10 to 1, so agents are aimed at the queue of handoffs between specialists, not at engineering judgment.
  • 03A model can match conveyor dimensions and machine motion and still assume perfect part supply, so any green label should be tied to the specific question it was validated for.

Nov 19, 2026

Why Atlas Copco's cobot on a Mexican transmission line paid back in 10.5 months

Why Atlas Copco's cobot on a Mexican transmission line paid back in 10.5 months

Mexico's industrial robot installations fell 4% in 2024 and 7% in 2025 while global installs rose 11%. A key theme in Valverde’s analysis is that robots tend to pay back where utilization is high—especially in bottleneck stations running across multiple shifts—though support labor and other limits can erode gains. Atlas Copco reports a 10.53-month payback for a cobot on a three-shift transmission station, a supplier-reported result.

  • 01A first-year ROI of 13.92% and a 10.53-month payback describe nearly the same fact: if savings arrive evenly, a station that pays back inside a year returns its cost plus about 14% in its first 12 months.
  • 02In Mexico’s robot market, application and utilization impact payback more than brand, with high-utilization bottlenecks being potential spots for automation if they truly constrain output across shifts and gains are not offset by support labor costs.
  • 03Mexico's automotive share of robot installs was 63% in 2024; a falling share with steady total installs could indicate automation is spreading to other industries.

Nov 19, 2026

Explore More Industrial IoT Insights

Read more expert perspectives from across Industrial IoT.

Browse Industrial IoT Hub

About the Expert

MarketScale Newsroom
MarketScale Newsroom

Editorial Team

MarketScale

The MarketScale Newsroom reports on the companies, technologies, and trends shaping 16 B2B industries. It turns primary sources and expert commentary into clear, useful coverage for the people doing the work.

For B2B teams

Your experts could be publishing here

Stories like this one run on content MarketScale captures from real practitioners. See how your team's expertise becomes coverage in Industrial IoT and beyond.

Book a Demo

Or call us. No forms required. We pick up. 214-945-2512