One, maybe two — and only two once the first is proven and someone owns each properly. Seed-stage teams typically don't have the budget, headcount, or data volume to run four channels well simultaneously; spreading thin usually means every channel gets a mediocre, under-resourced attempt instead of one channel getting a real test. The evidence you need — is this channel actually producing qualified pipeline at a workable cost — takes focused weeks of iteration to surface, not a scattershot.
Add a second channel once the first has a repeatable, understood process, not before — Kando typically has new clients prove one channel fully before green-lighting a second, for exactly this reason.