Morgan Stanley cut its riskiest reconciliation job in half — by making its agents less autonomous

Morgan Stanley’s internal FIXR system helps controllers reconcile profit-and-loss breaks across trading systems, cutting work for some books from up to six hours to two or three. The bank says the system saves about 1,500 hours per week by proposing resolutions while humans remain closely involved and convert decisions into repeatable rules.

The deployment shows enterprise AI moving beyond coding and customer support into accuracy-critical financial operations. Morgan Stanley’s approach suggests human-in-the-loop agent systems may gain trust faster in regulated industries than fully autonomous workflows.