Hand tools still beat machines for one-offs
Setup time, and where automation stops paying for itself.
Hand tools still beat machines for a single part because a machine's advantage only shows up once its setup cost is spread across many identical pieces, and a one-off part never gets the chance to spread that cost across anything at all, so the machine ends up paying a large fixed penalty to produce exactly one thing a hand tool could have produced without paying it.
The short version
Every machine that removes material faster than a hand tool does so by trading a slower start for a faster middle. A milling machine has to be aligned, a workpiece clamped square, a program written or a fence set, and none of that time cuts any metal at all, it only prepares the machine to cut metal quickly once it begins. A hand file needs none of that preparation, it can begin removing material within seconds of being picked up, which means that for a single part the comparison is never fast machine against slow hand tool, it is slow-to-start machine against instantly-starting hand tool, and for exactly one part the head start rarely gets overtaken before the job is finished.
This is not a criticism of the machine, whose entire design assumes it will be asked to repeat the same cut many times over. A jig that takes an hour to build and align is a poor use of that hour if only one part will ever be cut in it, but it is an excellent use of the hour if the same setup will go on to produce the fortieth part just as accurately as the first. The mistake is applying the machine's logic to a job the logic was never built for, treating a single part the same way a production run would be treated and paying for repeatability that a one-off part has no use for.
The tailored-hem comparison
Anyone who has hemmed a single pair of trousers by hand rather than digging out a sewing machine for the job already understands this trade-off from the other side of the workshop. Threading a needle and running a line of stitches by hand takes only a few minutes and can begin the moment the trousers are picked up. Setting up a sewing machine for the same single hem means finding it, plugging it in, threading both the top and the bobbin, adjusting the tension, and running a test line on scrap fabric first, easily ten minutes of preparation before the machine has sewn a single real stitch. The machine sews faster once it is running, sometimes remarkably faster, but for one hem that speed never gets the chance to matter, because the job is already finished by hand before the machine would have been ready to start.
Where the crossover actually sits
The crossover point, the number of identical parts at which the machine's setup cost has finally been paid back by its higher cutting speed, depends entirely on how long the setup takes relative to how much faster the machine cuts once running, rather than being any kind of fixed number. A five-minute setup on a machine that cuts ten times faster than a hand tool pays for itself very quickly, often on the second or third part, while a two-hour setup on a machine only marginally faster than a skilled hand barely pays for itself even across a production run of fifty. This is why the same shop that owns a milling machine, a lathe and a bench of hand tools will often reach for a file on a genuine one-off and only reach for the machine once a drawing calls for more than a handful of identical copies.
The same arithmetic explains a pattern that looks strange from outside a workshop, where an experienced machinist will sometimes glance at a drawing, count the number of parts needed, and reach for a hacksaw and a file rather than walking over to a perfectly good mill sitting idle three metres away. That decision is not laziness or nostalgia, it is the crossover calculation done instantly and from memory, weighing the setup time the machine would demand against the small number of parts actually on order, and concluding correctly that the file will be finished before the machine is even aligned.
One figure worth keeping in mind
A basic CNC setup, fixturing a workpiece, touching off the tool and proving the program on air before it is trusted to cut, commonly takes longer than the machine then spends actually cutting a single simple part, sometimes several times longer. A hand tool spends essentially all of its time on the cut itself and almost none on preparation, so the two approaches are not really being compared on speed at all, they are being compared on how much of the total time each one wastes before the cutting even starts.
What follows from this
Recognising setup time as the real variable changes how a one-off job gets planned, because the question stops being which tool cuts fastest and becomes how many parts this particular setup will actually need to produce before it is worth paying for. A prototype needing one bracket is usually a hand-tool job even in a shop full of capable machines, while the same bracket needed in a production run of five hundred is a machine job even if the machine is slower to set up than anyone would like, because five hundred repetitions is more than enough to absorb almost any fixed cost the setup demands.
Where this stops being true
The advantage hand tools hold on a single part collapses the moment tolerance, not speed, becomes the deciding factor, since a machine holds a position it was set to far more reliably than a hand holds a file steady, and a part that genuinely needs a machine's accuracy is worth the setup time regardless of how few copies are being made. Setup time decides the choice only among methods that can actually hit the tolerance the drawing calls for, and once a hand tool cannot, the comparison in this article stops applying entirely.