Morgan Stanley AI reconciliation slashes P&L processing time by half

Morgan Stanley AI reconciliation: a new model for finance automation

Morgan Stanley AI reconciliation has slashed the time needed for profit‑and‑loss (P&L) reconciliation from up to ↓ 6 hrs per book to an average of 2‑3 hours, delivering a weekly saving of ↑ 1,500 hrs across roughly 100 controllers worldwide.

The internal system, dubbed FIXR, operates as a human‑in‑the‑loop platform: agents surface mismatches, propose fixes based on learned rules, and defer final approval to senior staff. Each recommendation is logged, reviewed, and fed back into the model, turning ad‑hoc decisions into repeatable logic.

“It’s more a co‑worker than a copilot,” Todd Johnson, Managing Director at Morgan Stanley, told Reuters.

FIXR orchestrates three specialized agents: one that translates historical guidance into start‑of‑day actions, another that captures controller behavior into rule sets, and a third that codifies recurring patterns into durable automation. The approach mirrors findings from a recent Bloomberg survey highlighting the need for process‑first AI deployments.

By limiting the model’s discretionary scope, the firm maintains deterministic control while still harnessing generative insights where appropriate. Governance remains anchored in continuous human accountability, ensuring that every automated clearance is traceable and auditable.

Analysis by Zara Blackwood (Rapid Response Intelligence Analyst).