Quality has always been a competitive moat. AI is changing how it's built.

In the 2000s, US medical device revenue doubled, driven by more products and increasingly complex ones. But adverse event reports grew even faster, outpacing revenue by almost 8% annually.

Compliance ≠ quality

In response, the FDA conducted an in-depth analysis of device quality data and formalized what the industry already knew: compliance and quality are not one and the same. Maintaining compliance with FDA regulations didn't ensure good product quality. Companies could pass inspections and still produce poor devices. And they could fall short of regulatory standards and still make excellent products.

In 2011, the FDA launched the Case for Quality to transform its focus from regulatory compliance alone to quality, reframing compliance as the floor, not the ceiling. Accompanying its launch, the FDA's analysis showed that manufacturers focused on quality outperformed their peers with more productivity, fewer complaints, fewer CAPAs, and lower costs. On the flip side, the average drop in share price following a major quality event between 2006 and 2009 was 9.8%.

Even then, the case for quality was already clear: compliance is just a baseline. Quality is a strategy that drives profit, customer satisfaction, and competitive advantage.

Quality by design, not by inspection

In 2014, the FDA worked with manufacturers, health care providers, patients, payers, investors, and the Medical Device Innovation Consortium to form a more transparent and collaborative approach to evaluating — and promoting — quality. The framework they developed borrowed from the defense and software industries and adapted it for medtech: the Capability Maturity Model Integration (CMMI). CMMI replaced pass/fail inspections with a spectrum that measures systems, not snapshots, shifting the goal from clearing an inspection to building a quality system that predictably produces good outcomes.

In 2018, the FDA piloted this approach in what is now the permanent Voluntary Improvement Program (VIP). Under it, a certified third party evaluates your quality system through site visits, system demonstrations, and conversations with staff. That evidence is then mapped against the CMMI model as an external benchmark of best practices that go beyond regulatory minimums.

Critically, your appraisal results give the FDA much deeper visibility into your quality system than a periodic inspection can, and that visibility translates directly into how the FDA works with you. Manufacturers who participate can trade routine inspection cycles for risk-based ones and see their submission review and approval times drop by 75-90%.

VIP isn't just about changing how you work with the FDA — the business results speak for themselves: $15M increases in product sales, 65% increases in daily production, and 95% reductions in complaints.

How AI is changing the quality game

Now, AI is putting quality — and the systems built around it — to a new kind of test. The FDA's list of cleared AI-enabled medical devices (AIMDs) has more than doubled since 2022, with clearances now approaching 1,500.

A research letter from authors at Johns Hopkins, Yale, and Georgetown studied 950 FDA-cleared AIMDs and found that 60 had been associated with 182 recall events — 43% of which occurred within 12 months of clearance, roughly double the rate for non-AI devices. Because 510(k) clearance doesn't always require prospective human testing, many AI-enabled devices reach patients before anyone has systematically observed how the algorithm performs in actual clinical use. Unlike hardware, AI systems can degrade as the data they encounter in the real world diverges from what they were trained on. Devices without clinical validation had 2.8x higher odds of recall.

AI introduces a different category of quality failure than VIP was built to address. The old risks were bounded — manufacturing defects, supplier failures, design flaws with more easily identifiable root causes. The new ones don't exist until the device is in use: they emerge from the interaction between the algorithm, the real-world data it encounters, and the clinical environment in ways that can't be fully anticipated before deployment. Algorithmic drift, human-AI handoff failures, software updates that ripple across an entire system at once. A quality failure in an AI-enabled device reaches further and faster than in a purely physical one.

Rebuilding quality for the AI era

Even though AI is complicating the case for quality, the underlying logic holds — and applies more urgently than ever. Quality has always been a moat. When Baxter had to suspend sales of its Novum IQ infusion pump in July 2025, its stock fell 22.4% in a single day. The pump had been linked to 79 injuries and two deaths — the latest in a pattern of software-driven Class I recalls that had already cost the company $44 million in remediation charges. For smaller companies, the stakes are even greater: an early quality finding can prevent commercialization in the first place.

AI is making recalls happen faster, failures harder to anticipate before launch, and software quality failures more far-reaching than physical ones. The companies that built quality infrastructure before AI arrived have a moat — for now. The companies that respond to this new threat quickly — embedding quality throughout the development lifecycle rather than leaving it to the end — will build the next one.