Potentially, something like generating high quality random numbers. This assumes a great deal about the fidelity of the qubits and the control software though.
There is much better, faster, cheaper hardware out there to generate random numbers. Using a quantum computer as an RNG is akin to buying an airplane and using it to blow-dry a carpet.
I mean, I'm not sure what else you'd literally do with it. Simulating the claimed 8 qubits would require not much classical processing power at all, unless there is something fundamental about the TRNG behavior/fidelity of qubits. Entanglement is interesting but what are you going to do other than simply demonstrate entanglement? I guess you could run Deutsch's with it. Fun. This seems like more of a teaching machine than anything.
Small qubits count machines could be useful for teaching about qubit operations. I can see several lab assignments based around these machine: measure the readout fidelity, measure the gate fidelity, measure T1 time, etc.