In an instance I'm vaguely familiar with, early in the War on Terrorism, US intelligence agencies, which rely very heavily on signals-based intelligence, especially the ability to intercept cables (including encrypted ones, thanks to a backdoor in Crypt AG's devices), telephone and satellite communications, radio/broadcast, and digital / computer communications ... were apparently stymied in trying to gather intelligence on Al Qaeda as that adversary avoided use of modern communications, relying strongly on personal messengers and physical media. Humint capabilities of penetrating cells or converting individuals were said to be underdeveloped.
Effectively, the NSA had excellent vision where Al Qaeda did not exist, and was blind where they did.
The quantity and low cost of sigint can often lead to a deprecation of humint. It's not sexy, it's slow, it's low-volume, it's expensive, it doesn't generate large support contracts with government suppliers.
But it can put precisely the right eyes and ears in precisely the right place in ways sigint often has difficulty in achieving.
Mind: this is based on open-sources discussion at the time. How much this reflected actuality vs. speculation or public campaigning / turf wars between intelligence camps (CIA tends to be more humint, NSA more sigint AFAIU) I'm not sure.
This is a much more recent article (I'm referring mostly to the 2000--2005 period), but makes similar points:
https://www.aei.org/research-products/report/enhanced-human-...