My company has maintained <60 second response times on our support 24/7 for a few years now, and although obviously it does require some level of additional staffing to cover peaks in demand, it's significantly less than you might imagine - I'd say maybe 1.5-1.7x.
We did optimise this by filling the lower demand periods with other tasks which can comfortably be completed more slowly - eg reviewing identity documents for KYC, etc.
I see it as being that the overall bandwidth you need being pretty close to the same, just choosing to have very low latency for specific tasks, and making sure constant context switching is manageable.