You mean using the low plate voltage, saturation region in the same way that the same region is used in FETs (triode region)?
The traditional approach is either a variable Gm tube (which inevitably introduces some amount of signal distortion for a given SNR), or a dual-grid tube where the plate current is the product of the two grid transconductances (thus forming a one-quadrant multiplier). Examples being remote-cutoff pentodes like 6BA6, dual-control pentodes like 6HZ6 or 6HS8, beam control types like 6BN6, and pentagrid converters like 6BE6 (which I think is also remote cutoff? a number of pentagrid variants were made with both characteristics). There were also specialty tubes for more in-depth operation (like the 7360? beam deflection tube for two-quadrant multiplication).
Plate characteristics tended not to be used because of the weak current flow around zero volts, the poor control of it (triodes act more like Thevenin equivalent sources with a dependent voltage, moreso than a variable resistance; the plate resistance does still vary, of course), and the ponderous screen current required in the case of pentodes (total cathode current varies little with respect to plate voltage; in plate saturation, all the current that would've reached the plate, ends up turning back to the screen instead, and toaster-grid ensues).
Similarly, JFETs stink in that the triode region's width is Vgs - Vpo, and the slope is only variable over a modest, say factor of 2 or so range, for a reasonable drain voltage range (before entering linear (current saturation) mode). It's usable, it's certainly been used in practice before; but the wider range, more general mechanism is usually an OTA, or Gilbert cell or the like.
Tim