If AI agents start managing customers’ money, the risk is not only what one bot can do, but what many might do at once. Infosys financial-services executive Balmukund Shukla argues that banks need explicit controls for agent identity, authority and the speed at which money can move.
Watch Desk analysis
What happened
In a Forbes Councils opinion article published on 16 November, Shukla sets out how AI agents might monitor balances, compare rates, optimise payments and, with permission, initiate transactions. He warns that agents acting on similar signals could move deposits quickly and in the same direction, even when each individual decision makes sense for its owner.
His proposed response is a “Know Your Agent” framework alongside existing customer checks: identify the agent and provider, record who authorised it, define what it can do and for how long, and make that authority revocable. He also suggests graduated permissions, from observing and recommending through to initiating and executing transactions.
Why it matters
A bank cannot treat an agent with permission to read a balance as equivalent to one allowed to move substantial sums. Shukla argues that controls should scale with the agent’s authority, including stronger authentication and human intervention for higher-consequence actions.
There is a second, less individual problem. If many agents respond to the same rate difference or market signal, banks may face faster, more correlated withdrawals. Shukla says liquidity stress tests should account for that possibility, which he calls algorithmic liquidity risk.
Our read
This is a proposal for a possible shift in banking, not evidence that autonomous agents are already moving deposits at scale. But the design questions are refreshingly concrete: who is acting, what can it do, and how does a customer stop it? “The customer gave it my password” is not a particularly reassuring answer.
What to watch
- Whether banks build explicit, revocable permissions for AI agents rather than relying on shared customer credentials.
- Whether stress tests begin to account for rapid, correlated withdrawals prompted by automated decisions.
- Whether customers get a clear way to see what an agent can access, what it has done and how to stop it.
Discussion spark: Should banks let customers delegate financial decisions to AI agents under strict, revocable limits, or should moving money always require a fresh human approval?
Sources and evidence
- When AI Starts Moving The Money: Preparing Banks For The Agentic Era – Forbes (16 November 2026, 12:15 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.