AI “colleagues” move from pilot to production inside the big banks

Artificial-intelligence agents are no longer experimental novelties in finance: they are clocking in, collecting work tickets and handing finished tasks to human supervisors.

BNY Mellon formalises its digital workforce

Bank of New York Mellon now operates several dozen AI-based “digital employees.” Each bot receives a unique login to the firm’s internal systems, is placed under a named line manager, and is restricted to a single team’s data. Their early assignments focus on patching software vulnerabilities and double-checking payment instructions before funds move. Because they hold standard credentials, the agents can open the same coding tools and workflow dashboards as their human peers and submit fixes for approval without human prompting.

The bank’s technology group says new personas can be created in roughly three months, and that email and Microsoft Teams access is next. Tight scoping limits each AI’s data reach, minimising operational-risk exposure while still letting the agents act autonomously inside their lanes.

JPMorgan experiments with AI assistants at scale

JPMorgan Chase is taking a broader—but more tightly governed—approach. The firm already gives its 230,000 employees a general-purpose generative-AI chatbot and is evaluating “AI agents” that will be tailored to specific job families once questions around entitlements, audit trails and system integrations are settled.

Industry grapples with a new operating model

Banks experimenting with AI agents agree that the technology demands fresh control frameworks. Executives are weighing how much system permission a bot should hold, how to track its actions for compliance, and how human and digital staff will divide workloads. Cloud providers that supply many of the underlying models report that orchestration—rather than raw model accuracy—has become the main topic in board-level discussions with banks.

What comes next

Near-term roadmaps point to three priorities:

1. Broader task coverage – Banks plan to extend digital workers from narrow coding or payments checks into client onboarding, research synthesis and regulatory reporting.

2. Human-AI collaboration tools – Secure email, chat and task-assignment channels will allow bots to escalate edge cases directly to managers instead of halting workflows.

3. Governance blueprints – Institutions are drafting “digital-staff handbooks” that establish access tiers, performance metrics and incident-response procedures specific to non-human workers.

With deployment times measured in weeks and measurable productivity gains already reported, most global systemically important banks are expected to formalise AI head-counts within the next year. For employees and regulators alike, the key questions will shift from if AI belongs on the org chart to how it is supervised once it is there.

Photo Credit: DepositPhotos.com