Assembly
checked 1 Sep 2026

Machine tending

Task maturity · five steps · every value is an analyst assessment

Assembly

Early commercial

Putting parts together — insertions, fastening, connectors, clips — where the robot has to feel what it is doing rather than repeat a position.

Why we rate it this way

Screwdriving and press-fit cells run in production at scale in electronics and automotive, so this is genuinely commercial — but every one of those cells was engineered for its part set, and nothing transfers. The category has revenue without leverage: each new part is a fresh integration project, and the flexible-assembly demonstrations that would change that are still demonstrations. Early commercial until a cell absorbs a new part without an engineer.

What happens today

The incumbent process

Manual assembly at a bench or a moving line, with jigs, torque drivers and a trained operator absorbing every part variation without anyone writing it down.

Human hands remain extraordinarily good at compliant insertion and at silently compensating for parts that are out of spec — which is exactly what makes the automation gap so persistent.

How do robots attack it

Robotics approaches

Force-controlled robotic assemblyEarly commercial

An arm with force-torque sensing feeling its way through insertions and fastening, instead of relying on position alone.

example vendorsDoosan RoboticsFANUC
Caged arm tending cellMature

A fenced industrial arm serving one or more machines at full speed, with a guarded gate for operator access.

What does it demand

Technical requirements

Compliant insertion under tolerance stack-upANALYSTHandling flexible partsANALYSTError recovery without a humanANALYST
Who is doing it

Vendors and deployments

vendorsFANUC
Who buys it

Target customers

Appliance and consumer goods manufacturers

Medium-volume durable goods assembly is the segment humanoid and flexible-cell vendors most often name as their entry point.

Automotive component suppliers

Component subassembly runs at volumes and programme lengths that justify a dedicated cell.

Electronics manufacturers and EMS providers

High volumes, small parts and severe labour cost pressure make this the deepest installed base of automated assembly.

What is in the way

Adoption barriers

Engineering cost per product

Nothing transfers between cells. Each product means new fixtures, new grippers and new tuned search strategies, so the vendor's cost scales with customers rather than falling with volume. That is why the category has revenue but no compounding.

Product design that assumes hands

Products designed for human assembly hide fasteners, use floppy parts and demand two-handed coordination. Design-for-automation fixes it, but the design decision is made years before and by a different team, so the robot inherits a product that was never meant for it.

Line changeover risk

A cell that cannot be re-tasked in a model-year change is capital stranded on a product with a two-year life. Plants price that risk in, and it often outweighs the labour saving.

Does it pay

Economics

The honest unit of account here is engineering hours per part number, not robots per line; a vendor whose deployment cost does not fall with the tenth customer does not have a product yet.

Assembly vendors are routinely valued on installed cells, which flatters a business whose margin is consumed by per-customer integration. The metric that predicts whether the category compounds is the trend in engineering hours per deployment.