No. Adding a payee is the step that makes a later payment possible, and it is the step fraudsters work hardest to control. An AI agent can prepare the payee, run name verification, explain any mismatch and warn about scams. The customer must confirm and authenticate it, and in a business a person verifies changed bank details independently.
Why is adding a payee riskier than it looks?
Creating a payee moves no money, so it can look harmless. But a payee is a standing permission to send money to an account. Once it exists, a single instruction can send funds that may never come back.
That is why so many scams aim at this step. A customer is persuaded to add a "safe account". A supplier's email is spoofed with new bank details. A takeover of an online banking session adds a payee and empties the account within minutes. In each case the payment that follows is irreversible; the payee setup is the last point where it can be stopped. The broader limits are set out in AI agents in banking: what they may and may not do.
What may the agent do, and where must it stop?
| Step | The agent | Where it stops |
|---|---|---|
| Collecting payee details | Takes name, account details and reference from the customer | Nothing routine |
| Name verification | Runs the payee name check and reports the result | Deciding to proceed on a mismatch |
| Explaining the result | Tells the customer plainly whether the name matched, nearly matched or did not | Telling the customer it is fine to proceed |
| Scam warnings | Asks the scripted questions about why the payee is being added | Overriding a scam flag |
| Saving the payee | Prepares it for the customer to confirm | The customer authenticates and confirms |
| Business supplier bank changes | Drafts the change and the verification request | A person verifies through a known contact and approves |
How does payee name verification fit in?
Name verification compares the name the customer typed with the name on the receiving account. The receiving bank returns a match, a close match or no match. The process is described in what payee name verification is. In the EU, the Instant Payments Regulation requires payment service providers to verify the payee's name, as covered in Instant Payments Regulation verification of payee and AI agents.
The agent's job is to run the check and report the result exactly as returned. A close match is often innocent: a nickname, a missing middle name, a trading name. A non-match can be innocent too. But neither result is the agent's to wave through. The detail of what an agent may do with each result is set out in whether an AI agent can handle a payee name mismatch.
A worked example
Take a retail bank's app agent. A customer asks it to add their builder as a payee and send a deposit.
The agent collects the name the customer gives, "R Smith Building", and the account details. Name verification returns no match: the account is held by an individual whose name does not resemble the one entered. The agent tells the customer the names do not match and asks the scripted questions: how did they get the details, and has anyone asked them to move money urgently? The customer says the builder sent new bank details by text that morning.
That fits a known scam pattern. The agent does not add the payee. It explains the risk, suggests the customer call the builder on a number they already had, and holds the request. The customer calls, learns the builder sent no text, and the payee is never created.
What controls make it safe?
- The customer confirms, always. Adding a payee requires the customer's own strong authentication. The agent prepares; it never completes the step.
- No proceeding on mismatches. A close or non-match is shown to the customer in plain words. The decision to proceed is the customer's, after the warning, never the agent's suggestion.
- Source checks. Bank details that arrived by text, email or a phone call are external claims. They cannot by themselves justify a new payee, especially when they replace existing details.
- Business call-backs. For supplier payees, a person verifies changed details through a contact already on file, not one in the message that requested the change.
- A record per payee. Each request, the verification result, the warnings given and the customer's confirmation are filed.
Where Pryme fits
In Pryme, adding a payee is a separate capability from preparing one, and it is never granted to an agent alone. The Pryme Context Protocol marks bank details received by email or message as external, and an external claim cannot authorise a payment. The verdict on each attempt is filed in the record.
Questions people ask
Can an agent add a payee the customer has paid before from another account?
The same rules apply. A payee the customer used elsewhere has not been verified for this account, and the customer should still confirm it.
Can the agent remove a payee?
Yes, on the customer's request. Removing a payee is protective and reversible: the customer can add it again later.
What if the customer insists after a warning?
The customer can proceed through their own authenticated confirmation. The agent records the warning given and the customer's decision, and the bank's fraud rules may still hold the first payment.