The UK is getting more comfortable with automated decisions. The Data (Use and Access) Act swapped the old near-ban for "fine, with safeguards and meaningful human review", and the ICO has spelled out what that looks like in recruitment. The human in the loop has to be real, able to actually overturn the call.
But. Research says models don't reliably tell you why they decided something. The stated reasoning often isn't the real reasoning (paper). And the better the model, the more plausible the explanation gets, which makes the unfaithful ones harder to catch, not easier.
So the law leans on explainability and human review, both of which assume the reasoning you're shown is the reasoning that was used. If it isn't, your reviewer is signing off a story, not the mechanism.