NEW YORK–As BNY expands its use of artificial intelligence agents to handle banking operations, the company is applying a familiar approach to overseeing their work: assigning supervisors, monitoring performance and requiring human reviews of certain decisions.
According to a case study published by Microsoft, BNY has deployed nearly 140 so-called “digital employees” throughout the organization. These AI systems have login credentials, avatars, employee numbers in the corporate directory and human supervisors responsible for assigning tasks and reviewing their output.
The approach is part of a broader effort by the bank to introduce AI into its operations while maintaining the controls required of a highly regulated financial institution that processes approximately $2.5 trillion in transactions daily.

How BNY Reviews AI Performance
According to Microsoft, BNY treats oversight of its digital employees much like a traditional employee performance-review process.
Human managers use live dashboards to monitor several measures, including:
- Transaction volumes: How much work each AI agent processes.
- Processing times: How quickly assignments are completed.
- Missed transactions: Whether agents fail to process items requiring attention.
- Approval outcomes: Whether human reviewers accept or reject work completed by AI.
When problems emerge, supervisors can adjust the systems and retrain them.
Microsoft cited an example involving BNY’s payments operations in which human reviewers returned some AI-generated responses even though the responses were technically correct. The problem was that reviewers were accustomed to older procedures.
Rather than abandoning the technology, BNY addressed the issue through coaching, recalibration and retraining, similar to how managers might work with a new employee.
Human Approval Remains Part of Process
Microsoft reported that BNY initially assigned one digital employee a narrowly defined task involving international payments.
The agent examines vendor addresses, consults a mapping application to determine the correct country, validates country codes and submits corrected payments for human review.
According to the company, validations that previously required five or six minutes can now be completed in less than 30 seconds. Open payment investigations have declined by nearly 80%.
Another AI agent handles customer inquiries about payments. It reads incoming messages, retrieves information from multiple banking systems and drafts responses, which are then submitted to a human manager for review.
Microsoft said inquiries that previously took days to resolve can now be handled the same day.
Formal Oversight Before Deployment
BNY also subjects new AI applications to a governance process that Microsoft described as involving 16 steps, including review boards, compliance checks and model-risk evaluations.
The objective is to ensure AI systems can be monitored, controlled and audited before they are deployed.
The bank is also developing what it calls a Digital Employee Agency, which will formalize decisions about when AI agents should be used instead of human employees or contractors.
That process is expected to include onboarding, performance scorecards and procedures for retiring digital employees when newer technology becomes available.
Training Human Employees
BNY’s AI expansion includes a 40-hour employee training program that has graduated approximately 2,300 workers since launching in April 2025, according to Microsoft.

Participants must develop a working AI prototype addressing a business problem.
The bank has also invested in Microsoft’s Azure cloud infrastructure and Copilot technology, with approximately half its workforce licensed to use Microsoft 365 Copilot.
Michael Demissie, BNY’s head of applied AI, said the bank’s objective extends beyond automating repetitive assignments.
“AI is changing how work gets done. We’re investing in our people to do more with AI—not merely less of the mundane, but more of everything,” Demissie said.
Microsoft reported that BNY plans to continue expanding its digital workforce while shifting human employees toward more analytical, strategic and supervisory responsibilities.




