Automation to Autonomous Process Units: The Next Shift in Medical Device Manufacturing

Medical device recalls hit a four-year high in 2024. 

Class I recalls, the category covering the most serious risks, reached their highest level in 15 years, according to AdvaMed’s summary of Sedgwick’s 2025 U.S. State of the Nation Recall Index.¹ For the first time in over five years, device failure overtook process control error as the leading cause.

That’s a telling shift. Fewer problems get caught on the floor. More show up in the finished product, after it’s too late to fix at a lower cost.

Manufacturers are moving beyond incremental automation toward systems that respond to variations as they happen, instead of waiting for scheduled intervention. While the term “Autonomous Process Unit” isn’t yet an established industry classification, it offers a way to describe this new generation of manufacturing cells: production stations that continuously monitor their process parameters, correct drift on their own, and make routine inspection decisions in real time, all of which are inside limits a manufacturer has validated. 

Older equipment detects a problem and waits for someone. The newer generation fixes what it’s allowed to fix and interrupts a person only for the rest.

Where Today’s Manufacturing Model Breaks Down

Quality and operations leaders across device manufacturing tend to describe the same handful of problems, just in different words. A few surface more often than the rest.

The Skilled Labor Gap

Experienced quality staff is retiring faster than replacements can train. According to ETQ’s Pulse of Quality in Manufacturing 2025 survey,² 70% of manufacturers say labor shortages are impacting operations.

Disconnected Systems

Different teams built or implemented Manufacturing Execution System (MES), Quality Management System (QMS), and Enterprise Resource Planning (ERP) platforms at different times for different jobs. Engineers spend hours stitching data together by hand because these systems don’t talk to each other.

Manual Inspection at Scale

Sampling-based inspection was developed around more predictable production environments and became less effective as product variation increases.

Rising Cost of Quality and Regulatory Load

Scrap, rework, and deviation investigations pull engineers away from the work that prevents defects. ISO 13485 and EU MDR add their weight. The same ETQ survey found that 88% of manufacturers say the labor shortage has hurt product quality.

From product launches to M&A and manufacturing trends, the MPO Source delivers the medtech news you need twice a week. Subscribe here!

Why Traditional Automation Isn’t Enough

Fixed automation runs whatever recipe an engineer gives it, with no way to adjust when incoming material shifts or conditions change mid-shift. Statistical process control helps, but only partway: a control chart shows that a process is drifting, but someone still must interpret it. A few gaps stand out:

• Fixed automation cannot respond to variation it wasn’t programmed to expect.
• Control charts detect drift but leave correction to a person.
• Deviation and Corrective and Preventive Action (CAPA) reviews add days or weeks between an event and a documented fix.

What Autonomous Process Units Change
This isn’t another layer of software bolted onto the line. It changes what happens the moment a process starts to drift.

Continuous Monitoring

Instead of a scheduled check every few hours, sensors watch temperature, pressure, torque, and vision-based defect signals for as long as a line runs.

Self-Correction within Limits

A drifting unit adjusts itself inside boundaries set during validation and logs the change.

Inline Quality Decisions

Inspection happens where the part sits on the line, instead of at a separate downstream station that delays the decision.

Escalation with Guardrails

The equipment decides nothing outside the approved range; a person does. That boundary keeps this from being unsupervised manufacturing.

The Foundation Is Already Emerging

Medical device manufacturers are beginning to add the capabilities that define more autonomous production systems. A case study published in the Journal of Manufacturing Systems³ examined the digitalization of a regulated medical device manufacturing process for ureteral stents, where legacy production equipment was connected to modern data systems to capture machine signals, monitor performance, and support predictive maintenance. The approach demonstrates how existing manufacturing assets can evolve from isolated automated machines into connected production systems that provide real-time visibility, identify potential issues earlier, and support more proactive quality decisions. While these systems represent an early stage of autonomy, they illustrate the transition from automation that simply executes instructions to manufacturing environments that continuously analyze conditions and respond to variations.

Where CDMOs Can Create Value

Much of this transformation is likely to appear first among contract development and manufacturing organizations (CDMOs), where manufacturers can spread investments across multiple OEM programs. CDMOs operate the manufacturing infrastructure that many OEMs rely on, and as those environments become more intelligent and self-correcting, the relationship begins to shift from capacity provider to strategic manufacturing partner.

Continuously Monitored Production

Rather than providing manufacturing capacity, CDMOs can differentiate themselves by offering continuously monitored production backed by validated process verification and richer operational data, tied to yield and deviation outcomes rather than capacity booked and lead time promised.

Predictive Equipment Maintenance

Manufacturers fold condition monitoring into the agreement, catching wear before it becomes downtime.

Supporting Batch Disposition

Inline inspection data informs batch disposition decisions, shortening the lag between production and the call to ship.

What to Actually Measure

None of this means much until it shows up somewhere leadership looks. A handful of numbers tend to tell the story:

• First-pass yield reveals how often a process produces conforming product without rework, making it a direct indicator of process stability.
• Deviation and CAPA cycle time show whether teams resolve problems faster, not merely flag them faster, which is the distinction this shift is built around.
• Right-first-time release rate reflects how much manual review still stands between a finished batch and the customer—how much trust the organization places in its controls.

Implementation Strategy: Buy, Build, or Hybrid?

As device manufacturers and CDMOs evaluate how to implement Autonomous Process Units, the decision comes down to one of three paths: buy, build, or hybrid.

Buy

Off-the-shelf inspection and monitoring platforms can accelerate deployment and reduce upfront investment. However, these systems may offer limited customization and typically carry ongoing licensing costs.

Build

Although developing a custom system in-house, or by a specialized solution provider, often means a longer deployment time and higher upfront investment, it lets manufacturers tailor the system to proprietary processes, maintain greater control, and reduce dependence on recurring platform fees.

Hybrid

This approach provides faster deployment, allowing the manufacturer to customize third-party products with AI features without becoming locked into a one-size-fits-all platform.

Where Human Oversight Still Belongs

An Autonomous Process Unit still needs a firm line between what it decides on its own and what needs a signature. A few things tend to hold that line in practice:

• Parts inside a proven tolerance band move forward on their own; anything borderline goes to a quality engineer, no exceptions.
• The FDA’s Quality Management System Regulation⁴ took effect Feb. 2, 2026, amending 21 CFR Part 820 to incorporate ISO 13485:2016 by reference, raising the bar for documented, risk-based processes across the medical device sector.
• These units sit at the seam between operational technology and IT systems of record, so they need the same auditability and access control as anything else feeding a regulatory decision.

The Real Competitive Line Is Moving

The shift traces a clear arc. Fixed automation ran a set recipe with no awareness beyond it. Connected automation has added visibility, allowing systems to share data without acting on what they saw. Autonomous Process Units represent the next point on that line: cells that don’t just see a problem but correct it, because inside limits have already proven safe. What follows is factory-wide orchestration, where units like these coordinate with each other instead of operating independently.

The competitive advantage is no longer only automation. It’s the ability to transform every manufacturing cell into a continuously adapting, validated quality system that catches a defect before it ever becomes a recall. For CDMOs and OEMs, that’s the real line between managing capacity and owning quality outright.

References

1. bit.ly/mposoftware09261
2. bit.ly/mposoftware09262
3. bit.ly/mposoftware09263
4. bit.ly/mposoftware09264

More from this author: Agentic AI as the Brain of Personalized Drug Delivery Devices

Anshu Raj is a director of operations at Chetu, where he oversees the AI and engineering R&D portfolios, including high-tech medical devices. Raj, who holds certifications in PMP, Agile, and NetSuite Foundation, focuses on mid-market companies that want to accelerate innovation through on-demand engineering teams.