We propose a novel influence-based adaptive sample borrowing approach to enhance treatment effect estimation in RCTs by effectively incorporating external controls.