Next Shift

Deburring Is the Last Job Nobody Wants. It's Next.

Every shop has the corner. A bench, a file, a flap wheel and a part that came off the machine with an edge nobody can ship. Deburring is the job almost every shop still does by hand, and the one fewer people want to do. That is exactly why it is next.

At Productive Robotics in Santa Barbara, freshly cut spur gears used to go to a bench to be deburred by hand. Drop one and it was done. "With our parts, they're just scrap as soon as they touch the floor," production manager Troy Kirby told Gear Technology in May 2026. His summary of the job was shorter: "People just don't like to deburr."

Today a seven-axis cobot holds each gear against a 320-grit Scotch-Brite wheel. Kirby put scrap at about 10 percent before and under 1 percent after, with throughput up a conservative two times, possibly closer to three. Productive Robotics builds cobots, so it had every reason to try. The reason it worked is the part worth understanding.

Why the Robots Skipped This Job

Robots took on welding and painting decades ago. Deburring held out, and the reason is in the edge itself. A burr is not the same size on every part. Neither is the part. A robot following a fixed path cannot feel contact the way an operator does, and it cannot adjust as an abrasive wears.

3M, in a white paper on robotic deburring, put the requirement plainly: a good candidate for robotic finishing is "a very consistent part in terms of burr size and shape." When parts vary, it can be hard for the robot to know when the part is done. Production Machining made the same point back in 2006: complex parts take significant programming time, and the programmed points may not match the real edge because the part itself varies.

So the job stayed with people. LaRoux Gillespie, author of SME's deburring handbook, wrote that roughly 73,000 U.S. workers deburred full time and that almost a million machinists and toolmakers did some deburring as part of the job. In a 2009 Cutting Tool Engineering column he put the worldwide cost of burrs at an estimated $27 billion to $50 billion a year. Those are his estimates, and they are not new, but they show the size of what never got automated.

The People Doing It Are Getting Harder to Find

The labor math has not gotten easier. The Manufacturing Institute and Deloitte estimated in April 2024 that U.S. manufacturing may need as many as 3.8 million additional workers between 2024 and 2033, and that up to 1.9 million of those jobs could go unfilled.

Deburring is also hard on the people who do it. Grinders and sanders vibrate, and long exposure can cause hand-arm vibration syndrome, which machinists know as white finger. NIOSH estimated in 1989 that 1.45 million American workers used vibrating tools, and that among the groups it studied, prevalence averaged about 50 percent. An ARM Institute project with Siemens, Yaskawa and Arconic described finishing work as done in dusty conditions with workers in uncomfortable positions.

Put those together and the job nobody wants is also the job that is hardest to staff. That is usually the moment automation shows up.

What Changed: Tools That Feel the Edge

The first fix was mechanical. Instead of asking the robot to be perfect, deburring tool makers gave the spindle some give.

ATI Industrial Automation introduced its Flexdeburr in 2004. It is pneumatically controlled so it can follow the part profile and compensate for surface irregularities, with about 8 mm of radial compliance, and ATI says that reduces robot programming. ATI also offers a version built for use inside a CNC machine. PushCorp takes it further with active force devices that use closed-loop force control; its AFD620 holds up to 115 lbf of force at plus or minus 0.4 lb resolution.

The point of all of it is the same. The tool absorbs the difference between the programmed path and the real edge, so the part does not have to be perfect and neither does the robot.

What Changed: Cobots Built for Finishing

The second fix was packaging. Finishing used to mean a custom robot cell. Now it can mean a kit. FerRobotics sells an Easy Finishing kit for Universal Robots cobots that combines its active contact flange, simulation software and Dynabrade tools. Productive Robotics paired its own cobot with a pneumatic deburring platform.

Machine tool builders moved the same way. Haas began selling its HC10 cobot and a cobot cell kit in November 2023, according to Plastics Machinery and Manufacturing. The HC10 handles 22 lb with a 39 inch reach, and the kit fits machines including the VF-1, VF-2SS, VF-2YT, DT-2 and UMC-500. That cobot is sold for machine tending, not deburring, but once a robot is standing next to the machine, the distance to a deburring station is short. Our Next Shift on the used-robot wave covers what those cells are starting to cost on the secondary market.

What Changed: Software That Plans the Path

The third fix is the one aimed at the hardest problem: programming a robot for a part you only run a few of. GrayMatter Robotics raised a $45 million Series B in June 2024 for robots that scan a part and plan their own sanding, polishing, buffing and grinding paths, with no hand programming. The company claims its systems run two to four times faster than manual work. In October 2025 it opened a 100,000 square foot headquarters in Carson, California, with more than 25 active robotic cells.

GrayMatter's pitch is surface finishing more than edge deburring, and its numbers are its own. But the direction is the same one our Next Shift on AI process planning described: the program writes itself from the part.

Resell CNC Take

The first deburring most shops automate will not involve a robot at all. It will be a brush or a back-side deburring tool in the spindle of a machine they already own. Start there, prove which edges are repeatable, and the parts left on the bench will tell you whether you need a cobot.

The Cheapest Robot Is the Spindle You Already Have

Not every automated deburr needs a robot. Some of it happens before the part ever leaves the machine.

Xebec's ceramic-fiber brushes run in the spindle like a cutting tool. MoldMaking Technology reported Xebec's claim that one mold-part test ran 57 times faster than manual methods, a vendor figure worth testing on your own parts. Heule's COFA tools deburr the front and back of a through-hole in one pass, without stopping or reversing the spindle. A burr that never leaves the machine never reaches the bench.

This is the part of the trend any shop can start on Monday. It costs a tool holder, a program and some spindle time, and it works on a used mill as well as a new one. Our Next Shift on done-in-one machining makes the same case for fewer setups: every operation the machine finishes is one less trip across the floor.

Where the Numbers Say We Are

Robot adoption is growing, but metalworking is not leading it. The International Federation of Robotics reported in September 2026 that the United States installed 38,400 industrial robots in 2025, up 12 percent, making it the second-largest market in the world after China. The metal and machinery industry ranked second among U.S. industries with 3,000 installations, but that number was down 15 percent.

A3, the North American automation association, counted 36,766 robots ordered in North America in 2025, up 6.6 percent, with non-automotive orders growing faster than automotive. Market researchers are projecting growth for robotic deburring and finishing in particular; Technavio forecast in 2024 that the robotic cutting, deburring and finishing market would grow by about $4.47 billion between 2024 and 2028. That is a forecast, not a count.

Read together, the numbers say metalworking shops have room to catch up, and the bench in the corner is one of the most obvious places to start.

What It Means for Your Floor

Deburr in the machine first. Brushes, chamfer tools and back-side deburring tools work on the mill you already run, used or new.

Sort your parts by repeatability. The edges that are the same every time are the ones a robot or an in-machine tool can take. The variable ones stay on the bench for now.

Buy machines that are ready for a cell. Probing, automatic doors and robot interfaces make the next step cheaper. We have a 2023 Haas VF-2SSYT listed with a cobot and Haas VF-2SS machines listed with robot cells.

Count the real cost of hand deburring. Labor hours, scrap from handling and the people you cannot hire belong in the math.

Design for fewer burrs. The cheapest burr to remove is the one the toolpath never made.

The Next Shift

Deburring held out because it needed a hand that could feel the edge. Compliant tools, cobot kits and path-planning software are giving the machine that feel, one repeatable edge at a time. The bench in the corner is not going away this year. It is getting smaller.

Frequently Asked Questions