Manufacturers didn't wake up this year excited about AI. They woke up to recalls, tariffs, and not enough experienced people to catch a bad part before it ships. That combination is pushing AI into quality inspection faster than almost anywhere else on the plant floor, and the reasons are worth understanding even if your shop is a fraction of the size of the companies in the data.
Where the numbers come from
Octave, an industrial software company, published its third annual Pulse of Quality in Manufacturing survey on June 3, 2026. The fieldwork was run by Censuswide, an independent research firm, in the first quarter of 2026. It polled 2,263 managers and directors responsible for quality, compliance and operations at manufacturers with 1,000 to 50,000+ employees across the U.S., the U.K. and Germany. Trade press, including Quality Digest, covered the same findings.
Two caveats up front, because they matter for how you read everything below.
This survey covers large manufacturers. A company with 1,000 employees is not a 30-person fabrication shop in Vernon, and the dollar figures in particular don't scale down neatly. I'll be explicit about that where it comes up.
Octave sells industrial and quality software, so it has a commercial reason to want AI-in-quality numbers moving up. Censuswide ran the fieldwork independently and the sample size and method are disclosed, but the sponsor's interest is worth keeping in mind.
One more note on the data itself. The release lists the labour-shortage figure as "up from 70% in 2026," which can't be right, since 2026 is the current survey year. I've used that figure only as this year's finding and left the year-over-year comparison out.
AI in quality jumped in one year
Forty-seven percent of the manufacturers surveyed said they're already using AI in quality processes, up from 33% in 2025. Another 43% plan to deploy it within two years.
That's a fast move for a function that has traditionally been conservative about new tools, and for good reason. Quality is where a mistake turns into a customer complaint, a warranty claim or a recall. People who run quality departments don't adopt things because they're new.
What they're actually using it for
Among manufacturers using AI in quality, the top uses were:
- Document automation, at 48%
- Training, at 46%
- Defect detection, at 44%
Look at those three together and you can see where the pressure actually sits. Paperwork. Catching bad parts. And getting new people up to speed fast enough that they can catch bad parts too.
I spent eighteen years in construction before this work, and the parts that failed inspection were never the ones anyone expected. They were the ones nobody had time to check twice. That's the gap defect detection is aimed at. It doesn't replace an inspector. It extends what one experienced person can reliably cover on a busy day.
Document automation leading the list shouldn't surprise anyone who's worked in a regulated or audited environment. Inspection records, nonconformance reports, certificates of conformance and corrective action paperwork consume a lot of skilled hours. That's the same pattern I wrote about in AI for engineering documentation: the expensive people end up spending their time on the record of the work rather than the work.
What one bad batch costs now
The recall numbers explain why quality suddenly has budget behind it.
Seventy-five percent of the manufacturers surveyed had a product recall in the past five years. Fifty-nine percent said a single recall now costs between $10 million and $49.9 million, up from 48% who said the same the year before. The most common cause, cited by 47%, was supply chain issues.
Your numbers aren't in that range. A small shop isn't carrying recall exposure in the tens of millions. But the failure that triggers a recall at a large plant is the same failure that costs a smaller shop a customer: a defect that should have been caught and wasn't.
A part that leaves your shop with a flaw doesn't cost you when it ships. It costs you when it comes back.
For a smaller manufacturer, "comes back" usually means rework at your expense, a credit on the next invoice, an urgent remake that bumps scheduled work, and a customer who starts inspecting your parts more closely or quietly sends the next order somewhere else. None of those show up as a single dramatic number. They add up anyway.
The supply chain finding is worth noting too. Nearly half of recalls traced back to supply chain issues, not the manufacturer's own process. Incoming inspection and supplier quality records are part of the quality picture, and they're often the least organized part.
Where defects slip through in a smaller shop
Picture a 30-person machine shop running a mix of repeat work and short custom runs. The repeat jobs are well understood: the setup sheets exist, the first-article inspection is routine, and the long-time machinists know what a bad part looks like before it reaches the bench. The custom runs are where trouble lives. New drawings, tight deadlines, a material substitution because the usual stock was late, and an inspection step that gets compressed because the truck leaves at three.
Nothing about that is careless. It's the normal pressure of a busy shop. But it's also exactly the situation where a defect gets through, and it's the situation the large manufacturers in this survey are trying to address at scale. Their version involves supply chain issues behind nearly half of recalls. A smaller shop's version is a late material substitution that nobody flagged to inspection.
Seeing that pattern in your own records, if the records exist, is usually more valuable than any tool. It tells you where the risk concentrates, which is where any additional checking, human or automated, should go first.
The labour shortage is a quality problem
This is the finding I'd pay most attention to.
Seventy-eight percent of manufacturers said labour or skills shortages are affecting their operations. Eighty-five percent said that shortage is hurting product quality directly, not just slowing production down.
The connection is worth spelling out. It isn't only that manufacturers can't find workers. It's that they can't find enough experienced ones, the people who catch by eye and by feel what a newer hire hasn't learned to look for yet. When that experience retires or moves on, the quality knowledge often leaves with it, because a lot of it was never written down.
AI in inspection doesn't solve a hiring problem. What it can do is change how much one experienced person can reliably cover while you're still trying to fill the roles you're short, and help newer staff get consistent faster. That's also why training shows up so high on the use-case list. The two problems are linked.
The part that still needs a person
There's a caution worth stating plainly. A defect-detection tool is only as good as the examples it learned from and the standard it's checking against. Somebody who understands the part, the customer's tolerance and the failure modes still has to define what "good" looks like and review what the tool flags. Put AI on an inspection station without that and you've added speed to a process nobody's actually supervising.
Questions to ask before AI goes near inspection
If you're considering any kind of AI in your quality process, a few plain questions will tell you a lot about whether you're ready.
- Is the acceptance standard for each part written down somewhere, or does it live in one inspector's head?
- Do you have examples of good and bad parts, with the defects labelled, that a tool could learn from or be checked against?
- When a defect gets through today, does anyone record where it was missed and why?
- Who reviews what the tool flags, and how much of their day would that take?
If most of those answers are "not really," the first improvement probably isn't AI. It's getting the inspection knowledge out of people's heads and into a form anyone, or any tool, could use. That work pays off on its own, even if no AI ever gets involved, because it's also what you'd need to train the next inspector you hire.
Quality got promoted
Two last numbers show how far this has moved. Sixty-three percent of the manufacturers surveyed now treat quality as a company-wide strategic initiative, up from 38%. And 71% expect to increase quality spending in 2026, up from 60% in 2025.
Quality used to live in the back office: a department that signed off on things and got blamed when something slipped through. In the companies in this survey, it's increasingly a leadership topic with its own budget line. Recalls, tariffs and thin crews are the reason, and those pressures aren't limited to large plants.
The same shift tends to change who gets involved. When quality is treated as a back-office function, the decisions about it get made by the quality team alone. When it becomes a strategic priority, purchasing, production, sales and leadership all start to have a stake in it, because recalls and returns hit their numbers too. That broader ownership is often more valuable than any tool, since many quality problems start in places the quality department doesn't control, like supplier selection or a rushed delivery promise made by sales.
What this means at shop scale
It's worth being direct about the gap between this survey and a small manufacturer. The respondents work at companies with at least a thousand employees, across three countries, with quality departments and compliance staff. Octave sells software to exactly that market. Taking their percentages as a target for a local shop would be a mistake.
What carries over is the underlying logic:
- Defects are getting more expensive to miss, at every scale.
- Experienced inspectors are harder to replace, and their knowledge is often undocumented.
- Quality paperwork eats skilled time that could go to catching problems.
For a smaller operation, the starting point is usually not an AI vision system. It's understanding where defects actually get through today, how inspection knowledge is captured or not captured, and how much time quality records consume. Those answers decide whether AI belongs in your inspection process at all, or somewhere else in the shop entirely. The manufacturing overview has more on how that tends to play out.
Working through those questions for your own operation, at your size rather than a plant twenty times bigger, is what the paid AI Discovery and Readiness Assessment is built to do. It includes a discovery call, a scored review of your workflows and data, and a written report on where quality investment would actually pay off.