To justify physical AI costs, choose high-value use cases

Industries adopting physical AI

Manufacturing and healthcare are among the more notable industries exploring and benefiting from physical AI to automate business operations, provide services and improve products.

Manufacturing

Applications such as machine tending, assembly, welding and inspection blend together AI, computer vision and physical AI embodiment.

Joshua Morley, group chief AI officer at digital engineering consultancy Akkodis, pointed to inspecting for microfractures and other defects in a heat-treatment process on a manufacturing line. Historically, the inspection might have been done with a fixed-position overhead camera. Physical AI, however, goes beyond that static point of view, with robotic arms providing mobility. A six-axis or collaborative arm typically carries a camera and its own lighting at the end-effector, with edge compute alongside, Morley said. "If there's something that might not be completely clear or poor lighting, he added, "it can maneuver itself into a better position, maybe zoom in, and manipulate the artifact."

The ability to detect defects more accurately improves quality. "The gain is consistency, coverage and traceability: 100% inspection rather than sampling, the same standard at the end of the shift as at the start and a digital record of every inspection," Morley said. The digital record, he noted, often matters more than the individual inspection pass/fail because it shows a heat-treatment process drifting before it produces scrap.

But quality-improvement claims are only credible when compared with a measured baseline of existing escape and false-reject rates, Morley cautioned. An escape occurs when a defective part makes it undetected through inspection, while a false-reject flags a good part as defective.

Overall, physical AI adoption in manufacturing is gaining momentum. The pace at which physical AI captures tangible business value is accelerating daily, said Mark Cumm, senior director and global aerospace lead at business and technology consultancy Slalom. "This is true for all manufacturers, regardless of size, product complexity or markets served," he said.

Benefits in the area of plant maintenance include heightened operational reliability, significant financial gains from market growth and lower costs, throughput and uptime gains, Cumm said. He noted that Slalom works with a top 25 global manufacturer that uses AI/ML on the edge and interfaces with IoT inspection devices such as digital cameras, lasers and sensors.

Cumm also pointed to product safety, quality and new revenue, collectively, as another notable area for physical AI implementation. He cited an example of a transportation products manufacturer using AI/ML, predictive analytics and integrating sensors to detect irregularities. The manufacturer also draws on decades of field experience in this space to optimize complex processes. That experience, paired with physical AI, ensures actionable insights to safeguard machinery uptime, with a digital feedback loop to design engineering for product improvements and new services, Cumm explained.

Healthcare

Physical AI shows the clearest near-term value in industries already strained by labor shortages, safety risks and operational complexity, said Blair Kerr, industry director at Slalom. "Healthcare is perhaps the most meaningful place to start, and the technology tends to land best when it supports care teams rather than replaces them," she said. "A few applications are already proving out."

Healthcare is perhaps the most meaningful place to start, and the technology tends to land best when it supports care teams rather than replaces them.

Kerr cited exoskeletons and robotic support as examples. Those applications help reduce the physical toll on nurses, who suffer strain injuries at a much higher rate than most professions.

Mobility and rehabilitation devices, meanwhile, give patients steadier and more iterative support during recovery, Kerr said. Wearables and connected monitoring tools catch problems earlier, often before they become emergencies and generate data critical to understanding a patient's entire condition, she added.

The growing use of AI-enabled real-world evidence is starting to change how care decisions are made at the bedside, Kerr said. Hospitals using such evidence are seeing gains in clinician productivity, fewer workplace injuries and stronger workforce retention -- burnout and physical strain remain two of the top reasons nurses leave the profession, she noted.

Building a business case for physical AI

Physical AI calls for a more expansive view of investment, said Anant Adya, executive vice president at IT services provider Infosys. "Physical AI makes it especially important to broaden how we consider ROI," he said. "Some of its most meaningful value may come from outcomes such as fewer defects, less downtime, faster inspections or reduced worker exposure to hazardous environments. Those benefits can be missed when organizations rely on a narrow set of financial metrics."

Some of its most meaningful value may come from … fewer defects, less downtime, faster inspections or reduced worker exposure to hazardous environments.

Physical AI systems, which require advanced AI chips and processors, remain more expensive than traditional industrial robots, despite mitigating factors such as component commoditization, according to Deloitte's "Tech Trends 2026" report. This cost difference puts the onus on decision-makers to build a solid case for investment.

Businesses should approach cost justification with a clearly defined operational problem and a tangible value at the end, Morley advised. That approach includes setting an objective, based on the operational problem, and determining how to measure the value realized.

Those steps apply to both digital AI and physical AI. But many businesses struggle with the process. For example, a business might focus on measuring productivity value, finding that it's 30% faster at completing tasks or able to complete X more tasks per working day, Morley explained. But the business could fail to make the leap from increasing output to achieving outcomes such as cost reduction or reassigning employees to more enriching work.

"They're not able to direct that productivity value into either business impact or human-centered value," Morley said. To bridge the gap, a business must decide which organizational changes to pursue with physical AI before it begins an implementation, he said. That could mean capacity redeployment or process re-engineering. Such transformation activities translate productivity value into impactful business value.

In the previously mentioned inspection use case involving microfractures and other defects, a 30% speed-up in the QA process could potentially enable one inspector to cover two manufacturing lines. The physical AI system provides a high-pass filter across everything, with the experienced inspector handling special cases. That dual coverage -- the transformation activity -- provides the "business impact," Morley explained. The benefits include increased coverage per inspector and cost avoidance as volume grows.

Alternatively, the company could shift the productivity value toward human-centered value by redesigning an employee's role, so the inspector covers one line while spending time performing root-cause analysis, Morley said.

More investment factors: utilization, scale, variance

Other considerations such as utilization and scale influence physical AI investment decisions. Businesses can weigh such factors by multiplying the system utilization rate (U), degree of automation (A) and value of the tasks automated (V), then subtracting the deployment cost (C):

(U × A × V) − C = ROI

Using this multi-value automaton ROI equation, the investment case collapses, regardless of the technology, if any of the three multipliers is low, Morley said. Utilization can be a sticking point, since most business cases assume utilization rather than measure it, he added.

Utilization is the crucial component. If it's such an intermittent job, or it's not something that can't be scaled, then the economics don't make sense.

"Utilization is the crucial component," he said. "If it's such an intermittent job, or it's not something that can't be scaled, then the economics don't make sense." Scale is important relative to upfront hardware expenses, the cost of integration and other fixed costs, but the consideration is scale of value, not scale of volume, he said.

Ways to measure value include time, such as physical AI performing more shifts and running unattended. That advantage is fairly straightforward, but other values add nuance to the cost justification. Task flexibility, for example, lets businesses redeploy one system across tasks or variants without reprogramming, Morley said. In this case, utilization increases without raising volume, which is where physical AI differs from fixed automation, he added.

But some use cases with limited variability might not be a fit for physical AI. "High-volume, low-variance physical automation probably doesn't need physical AI," according to the 2026 trend report "Physical AI Perceives, Reasons, and Acts in the Real World" by Forrester. "Conceptual models of the world, suites of sensors that perceive that world, and powerful computers to reason through possible alternative workflows cost a lot of money and complicate installation without necessarily making anything better."

Value density adds another wrinkle. A use case with high value density could justify investment even with limited utilization. "Low utilization," Morley explained, "still pays if what it prevents is expensive enough -- an unplanned outage or a person removed from a hazardous inspection."

John Moore is a freelance writer who has covered business and technology topics for 40 years. He focuses on enterprise IT strategy, AI adoption, data management and partner ecosystems.