Beyond the black box: operationalising explicability in AI for finance