For banks and credit unions
Agentic Payments for Banks and Credit Unions
Agentic payments are payments an AI agent prepares for a customer or member. The bank or credit union still holds the money, so it sets the limits, screens the payee and requires confirmation before anything moves. This page covers the institution side: accounts, ACH, domestic wires and internal transfers.
Read the agentic banking guide
What agentic payments are
Agentic payments are payments an AI agent prepares on behalf of a customer or member. The person asks. The agent drafts the payment. Rules outside the model check it. The person confirms the exact details. Then the payment can go out, and the whole path is logged.
The agent acts only when an authenticated customer or member asks and confirms the exact details.
Accounts, ACH, wires and transfers
Card-network programs center on a person paying a merchant, often with a card at checkout. This page is the institution view: deposit accounts, internal transfers, ACH and domestic wires.
The bank or credit union holds the money, so the bank or credit union holds the practical responsibility for limits, screening and confirmation. Segment guides: AI for community banks and agentic AI for credit unions.
Two ways a request can arrive
Both paths end in the same place: the institution authenticates the person, applies its own limits and requires confirmation of the exact payment.
The institution's own agent
The customer or member is signed in to the institution's digital banking and asks for a payment. The institution already knows who they are from that session.
An outside assistant
Some assistants people use elsewhere will ask to act on a bank account. Treat that request as untrusted until the institution authenticates the customer, applies its limits and gets confirmation of the exact details.
A chatbot and an agent
A chatbot can describe a cutoff time or a wire form. An agent prepares the payment and stops for confirmation. If confirmation is missing, the payment stays unsent.
| Chatbot | Banking agent |
|---|---|
| Answers, then hands the customer or member to a form, a page or a person | Finishes the task the customer or member asked for |
| Explains how to send a wire | Prepares the transfer and asks the customer or member to confirm the exact amount, accounts and payee |
| Keeps its rules in the prompt | Keeps rules in code the model cannot change |
| Speaks with a generic script | Speaks in the institution's voice |
What the agent can move, and what it leaves
Stay inside deposit-account payments the institution already offers. Anything that changes credit, the customer record or an international wire goes to a person.
What the agent can do
- Check balances, accounts and holds.
- Search and explain transactions, such as how much was spent on dining last month. The math is done by a calculator. The model does not guess the amount.
- Move money by internal transfer, ACH or domestic wire. The customer or member confirms the exact amount, accounts and payee before anything is sent.
- Manage payees and saved wire templates.
- Link an external account and set up automatic savings.
- Show spending trends and simple forecasts, always labeled as estimates.
What stays with a person
These requests are handed to a person.
- Loans, mortgages and credit cards.
- Disputes, lost or stolen cards, and fraud freezes.
- Opening or closing an account, and changing contact information.
- International wires.
- Investment, tax or legal advice.
What happens before money moves
A sound payment has the same beats every time. The order is the control. Skipping confirmation, or skipping screening on a new payee, is the gap.
An authenticated customer or member asks, inside the digital banking they already use.
The agent drafts the payment: amount, account, and payee or saved wire template. The amount is calculated, not guessed.
The rules layer checks per-transaction, daily and monthly limits. A new payee is screened for sanctions (OFAC) and account risk. If screening is unavailable, the payee is not added.
The customer or member confirms the exact details. Every payment needs this, at any amount. Larger payments also need a one-time code.
The payment goes out only after those checks. The conversation and the action are logged step by step.
Checks the institution owns
Because the institution holds the account, it sets the caps, the screening and the moment the agent must stop. A longer control discussion is in how to let AI agents move money safely.
Requests, then a rules layer
The agent can request an action. It never touches payment rails directly. A separate rules layer decides whether that action can proceed.
Hard limits
The institution sets per-transaction, daily and monthly caps. A payment over a hard limit is declined. Nothing overrides that limit.
Confirmation, then a step-up
Every payment, at any amount, needs explicit confirmation of the exact details. Larger payments also need a one-time code.
Payee screening
A new payee is screened for sanctions (OFAC) and account risk before that payee can be paid. If screening is unavailable, the payee is not added.
Only the tools that request needs
The agent receives only the tools that request needs. A tool it was not given cannot be called, no matter what anyone types.
Tricky instructions
Hidden or tricky instructions in a message or a document are blocked or treated as plain data, and flagged for the institution.
Separate data, full log
Each institution's data is kept separate. Every conversation and every action is logged step by step.
The institution stays in control
A dashboard shows activity and security alerts. Staff can edit spending policy, block a user and switch an agent off immediately. Staff can also try the agent in a sandbox with realistic data before it goes live.
When a payment looks wrong
A wrong amount, a new payee or a strange instruction should fail the check or the confirmation, and the attempt should be visible to the institution. Hidden instructions are flagged, not followed.
AI banking safety walks through a poisoned invoice, a vendor impersonator and a request that mixes a normal task with a hidden one.
If a payment has already gone out, responsibility depends on the facts and the rules for that payment type. Electronic fund transfers can raise questions under Regulation E. Ask counsel. This page does not decide liability.
Where to start
Start with transfers between accounts the customer or member already owns. Then payments to payees the institution already has on file. Add new payees only when sanctions screening fails closed. Keep international wires with a person.
- Own-account transfers, with confirmation of the exact amount.
- ACH and domestic wires to payees already on file.
- New payees, after OFAC and account-risk screening is in place.
- A sandbox trial with realistic data before customers or members see it.
Related guides
Common questions
What are agentic payments?
Agentic payments are payments an AI agent prepares for a customer or member. On the institution side that means accounts, internal transfers, ACH and domestic wires. The institution holds the funds, sets the limits, screens the payee and requires confirmation of the exact details before money moves. The wider category is agentic banking.
What is the difference between a banking chatbot and an AI agent?
A chatbot explains how to send a payment and then hands the person to a form or a banker. An AI agent prepares the transfer, ACH or domestic wire and asks the authenticated customer or member to confirm the exact amount, accounts and payee. Money moves only after that confirmation.
How is agentic payments different from agentic commerce?
Agentic commerce usually means an assistant buying from a merchant, often with a card at checkout. Agentic payments here means the bank or credit union side: money leaving a deposit account by internal transfer, ACH or domestic wire, under rules the institution enforces outside the model.
Is it safe to let AI move money? What stops it from sending the wrong amount?
The agent can request a payment. It never touches payment rails directly. A separate rules layer checks the amount against the limits the institution set, and the customer or member confirms the exact details. A payment over a hard limit is declined. Nothing overrides that limit. Transaction math is done by a calculator. A control list is in How to let AI agents move money safely.
How do AI agents handle OFAC and sanctions screening?
A new payee is screened for sanctions (OFAC) and account risk before that payee can be paid. If screening is unavailable, the payee is not added. The payment does not go out while screening is down.
Can an AI agent send a payment before the customer asks and confirms?
A controlled design declines that. The agent acts only when an authenticated customer or member asks. The rules layer checks limits and the payee. The customer or member confirms the exact details. Every step is logged, and the institution can switch the agent off immediately.
What happens if the AI makes a mistake or someone tries to trick it?
Confirmation of the exact details, hard limits and payee screening are what stop a bad payment. Hidden instructions in a message or a document are blocked or treated as plain data, and flagged. AI banking safety walks through cases such as a poisoned invoice, a vendor impersonator and a request that mixes a normal task with a hidden one.
Who is responsible if an AI-initiated payment is wrong?
It depends on the facts and on the rules that govern that account. Electronic fund transfers can raise questions under Regulation E. This page does not reach a legal conclusion. Ask counsel. What the institution can show is the request, the checks, the confirmation and the log.